A 50‐year‐old man presented to the emergency department with progressive shortness of breath for 6 months. He described a dry cough, left‐sided chest pain, malaise, night sweats, and a 15‐pound weight loss. The patient had never smoked cigarettes, but he had been exposed to asbestos and wood dust when working at a sawmill. His physical examination was remarkable for decreased breath sounds at the left lung base. The admission blood tests were within normal limits. Chest radiography and a computed tomography (CT) scan of the chest were performed (the CT scan is shown in Figures 1 and 2). The CT scan showed a left pleural effusion with subpleural and peribronchovascular nodules. Also demonstrated on the CT scan were bilateral hilar and mediastinal lymphadenopathies with faint central calcification. As the left‐sided pleural effusion was initially suspected to be malignant, a thoracentesis was performed, and it revealed an exudative effusion. The total white cell count in fluid was 2100/L (lymphocytes, 76%), and cultures for aerobic and anaerobic bacteria, acid fasting bacilli, and fungi were negative. Cytology was negative for malignant cells. On the basis of the findings in the lung parenchyma and the presence of mediastinal lymphadenopathies, fiber‐optic bronchoscopy with bronchoalveolar lavage, protected specimen brushing, transbronchial needle aspiration, and transbronchial biopsies were performed. Mediastinal lymph node cytology was negative for malignant cells, whereas transbronchial biopsies showed noncaseating granulomas (Figure 3). At that time, our differential diagnoses of noncaseating granulomas included mycobacterium infection (although this usually presents caseating granulomas), berylliosis, histoplasmosis, and sarcoidosis. The tuberculin skin test (purified protein derivative) and serology for human immunodeficiency virus were negative. Bronchoalveolar lavage and cultures of lung tissue biopsies as well as needle aspiration from mediastinal lymph nodes were negative for mycobacterial, fungal, and bacterial organisms. The beryllium lymphocyte proliferation test was normal. Serologic antibodies for Aspergillus, Blastomyces, Coccidioides, and Histoplasma were negative. The urinary Histoplasma antigen was negative as well. The Department of Infectious Diseases was consulted, and an empirical treatment for histoplasmosis with itraconazole was started on the basis of the residence of the patient and the presence of noncaseating granulomas. After 1 month of antifungal treatment, there was no significant improvement. Video‐assisted thoracoscopic surgery with pleural biopsy was performed because of persistent pleural effusion and concern about an underlying infectious or malignant process. Pleural biopsies showed noncaseating granulomas (Figure 4). Pleural fluid was sent for adenosine deaminase (17 U/L) and flow cytometry (CD4/CD8 2.71). Cultures and cytology remained negative. A diagnosis of stage 2 sarcoidosis with pleural involvement was made, and treatment with prednisone was started.

Figure 1

Figure 2

Figure 3

Figure 4

Discussion

The overall prevalence of pleural involvement in sarcoidosis is about 3%. Patients with pleural sarcoidosis tend to be between 30 and 50 years of age, in contrast to the usual presentation of sarcoidosis between 20 and 30 years of age. The most common forms of pleural involvement are pleural effusions, pneumothorax, pleural thickening, and pleural nodules.1 Most effusions are usually small or modest in size, with few reports describing massive effusions.2 Recurrent pleural and pericardial effusions due to sarcoidosis have been reported as well.3 The fluid is typically a lymphocytic exudate, and almost all cases describe a CD4 predominant lymphocytic effusion with CD4/CD8 ratios ranging from 2.35 to 8.6.1 The presence of bloody pleural effusions in sarcoidosis most likely represents the rupture of small vessels that are compressed or infiltrated by granulomas.4

The majority of patients with reported sarcoid pleural effusions have stage 2 disease. With the progression of parenchymal disease, the prevalence of pleural effusions decreases, whereas pleural thickening and pneumothorax increase.5 It is important to emphasize that 40% of pleural effusions in sarcoidosis may be due to other causes, such as tuberculosis and mesothelioma. Our patient was initially treated with itraconazole as histoplasmosis is most prevalent in the Central and Eastern United States, especially in Ohio River valleys, where this patient lived.

The prevalence of a pneumothorax in sarcoidosis is up to 4%.1 Pleural thickening can be demonstrated in 11% to 71% of patients with pleural sarcoidosis, and 10% to 20% of these cases have thickening without effusion. Detection of subpleural nodules and cysts has been possible since the introduction of high‐resolution CT scans. Their prevalence in sarcoidosis ranges from 22% to 76%, and they are often described as masses that correspond to nodules seen in both parietal and visceral surfaces. Hilar or mediastinal lymphadenopathy is present on CT in 47% to 94% of patients with sarcoidosis. Lymph node enlargement is usually bilateral, most commonly with right‐sided predominance. The most involved stations are the right lower paratracheal, right hilar, subcarinal, aortopulmonary window, and right interlobar stations. Nodal calcification is noted in 53% with eggshell calcification present in 9%. The enlargement of internal mammary and pericardial lymph nodes requires the exclusion of lymphoma.6

The management of pleural sarcoidosis should be individualized because a majority of these effusions resolve spontaneously in 1 to 3 months.5 There have been reports of resolution in 2 weeks with steroid therapy. Incomplete resolution of the pleural effusions with progression to chronic pleural thickening or a trapped lung has been reported. There is agreement that oral corticosteroid treatment should be considered in patients with severe persistent or progressively worsening respiratory symptoms or declining lung function. Severe symptoms can be considered as those that interfere with essential aspects of the patient's daily life.7 The initial dosage of oral prednisone recommended by the American Thoracic Society, the European Respiratory Society, and the World Association of Sarcoidosis and Other Granulomatous Disorders guidelines is 20 to 40 mg/day.8 Further evaluation is recommended after 1 to 3 months. If the patient responds, the dose should be reduced gradually to a maintenance dose, such as 5 to 15 mg/day of prednisolone. American Thoracic Society/European Respiratory Society/World Association of Sarcoidosis and Other Granulomatous Disorders guidelines advise treatment for at least 1 year. Immunosuppressive, cytotoxic, and immunomodulatory agents have been used to treat patients failing or experiencing adverse effects of steroids. Favorable responses have been reported with methotrexate, leflunomide, azathioprine, cyclophosphamide, chlorambucil, cyclosporine A, antimalarials, tumor necrosis factor inhibitors, and thalidomide. Because of potential serious toxicities associated with cyclophosphamide and chlorambucil, these agents are not recommended.9

Our patient presented with pleural sarcoidosis with a pleural effusion and nodules. Treatment with 20 mg of prednisone daily was started initially. Four weeks after discharge, he was still dyspneic and had persistent left pleural effusion. He also had gained a significant amount of weight and developed bilateral lower extremity edema; these were thought to be secondary to prednisone treatment. Steroids were subsequently tapered, and leflunomide was started. His symptoms improved dramatically after 1 month of treatment with leflunomide and steroids, and 3 months later, his pleural effusion had completely resolved.

A 50‐year‐old man presented to the emergency department with progressive shortness of breath for 6 months. He described a dry cough, left‐sided chest pain, malaise, night sweats, and a 15‐pound weight loss. The patient had never smoked cigarettes, but he had been exposed to asbestos and wood dust when working at a sawmill. His physical examination was remarkable for decreased breath sounds at the left lung base. The admission blood tests were within normal limits. Chest radiography and a computed tomography (CT) scan of the chest were performed (the CT scan is shown in Figures 1 and 2). The CT scan showed a left pleural effusion with subpleural and peribronchovascular nodules. Also demonstrated on the CT scan were bilateral hilar and mediastinal lymphadenopathies with faint central calcification. As the left‐sided pleural effusion was initially suspected to be malignant, a thoracentesis was performed, and it revealed an exudative effusion. The total white cell count in fluid was 2100/L (lymphocytes, 76%), and cultures for aerobic and anaerobic bacteria, acid fasting bacilli, and fungi were negative. Cytology was negative for malignant cells. On the basis of the findings in the lung parenchyma and the presence of mediastinal lymphadenopathies, fiber‐optic bronchoscopy with bronchoalveolar lavage, protected specimen brushing, transbronchial needle aspiration, and transbronchial biopsies were performed. Mediastinal lymph node cytology was negative for malignant cells, whereas transbronchial biopsies showed noncaseating granulomas (Figure 3). At that time, our differential diagnoses of noncaseating granulomas included mycobacterium infection (although this usually presents caseating granulomas), berylliosis, histoplasmosis, and sarcoidosis. The tuberculin skin test (purified protein derivative) and serology for human immunodeficiency virus were negative. Bronchoalveolar lavage and cultures of lung tissue biopsies as well as needle aspiration from mediastinal lymph nodes were negative for mycobacterial, fungal, and bacterial organisms. The beryllium lymphocyte proliferation test was normal. Serologic antibodies for Aspergillus, Blastomyces, Coccidioides, and Histoplasma were negative. The urinary Histoplasma antigen was negative as well. The Department of Infectious Diseases was consulted, and an empirical treatment for histoplasmosis with itraconazole was started on the basis of the residence of the patient and the presence of noncaseating granulomas. After 1 month of antifungal treatment, there was no significant improvement. Video‐assisted thoracoscopic surgery with pleural biopsy was performed because of persistent pleural effusion and concern about an underlying infectious or malignant process. Pleural biopsies showed noncaseating granulomas (Figure 4). Pleural fluid was sent for adenosine deaminase (17 U/L) and flow cytometry (CD4/CD8 2.71). Cultures and cytology remained negative. A diagnosis of stage 2 sarcoidosis with pleural involvement was made, and treatment with prednisone was started.

Figure 1

Figure 2

Figure 3

Figure 4

Discussion

The overall prevalence of pleural involvement in sarcoidosis is about 3%. Patients with pleural sarcoidosis tend to be between 30 and 50 years of age, in contrast to the usual presentation of sarcoidosis between 20 and 30 years of age. The most common forms of pleural involvement are pleural effusions, pneumothorax, pleural thickening, and pleural nodules.1 Most effusions are usually small or modest in size, with few reports describing massive effusions.2 Recurrent pleural and pericardial effusions due to sarcoidosis have been reported as well.3 The fluid is typically a lymphocytic exudate, and almost all cases describe a CD4 predominant lymphocytic effusion with CD4/CD8 ratios ranging from 2.35 to 8.6.1 The presence of bloody pleural effusions in sarcoidosis most likely represents the rupture of small vessels that are compressed or infiltrated by granulomas.4

The majority of patients with reported sarcoid pleural effusions have stage 2 disease. With the progression of parenchymal disease, the prevalence of pleural effusions decreases, whereas pleural thickening and pneumothorax increase.5 It is important to emphasize that 40% of pleural effusions in sarcoidosis may be due to other causes, such as tuberculosis and mesothelioma. Our patient was initially treated with itraconazole as histoplasmosis is most prevalent in the Central and Eastern United States, especially in Ohio River valleys, where this patient lived.

The prevalence of a pneumothorax in sarcoidosis is up to 4%.1 Pleural thickening can be demonstrated in 11% to 71% of patients with pleural sarcoidosis, and 10% to 20% of these cases have thickening without effusion. Detection of subpleural nodules and cysts has been possible since the introduction of high‐resolution CT scans. Their prevalence in sarcoidosis ranges from 22% to 76%, and they are often described as masses that correspond to nodules seen in both parietal and visceral surfaces. Hilar or mediastinal lymphadenopathy is present on CT in 47% to 94% of patients with sarcoidosis. Lymph node enlargement is usually bilateral, most commonly with right‐sided predominance. The most involved stations are the right lower paratracheal, right hilar, subcarinal, aortopulmonary window, and right interlobar stations. Nodal calcification is noted in 53% with eggshell calcification present in 9%. The enlargement of internal mammary and pericardial lymph nodes requires the exclusion of lymphoma.6

The management of pleural sarcoidosis should be individualized because a majority of these effusions resolve spontaneously in 1 to 3 months.5 There have been reports of resolution in 2 weeks with steroid therapy. Incomplete resolution of the pleural effusions with progression to chronic pleural thickening or a trapped lung has been reported. There is agreement that oral corticosteroid treatment should be considered in patients with severe persistent or progressively worsening respiratory symptoms or declining lung function. Severe symptoms can be considered as those that interfere with essential aspects of the patient's daily life.7 The initial dosage of oral prednisone recommended by the American Thoracic Society, the European Respiratory Society, and the World Association of Sarcoidosis and Other Granulomatous Disorders guidelines is 20 to 40 mg/day.8 Further evaluation is recommended after 1 to 3 months. If the patient responds, the dose should be reduced gradually to a maintenance dose, such as 5 to 15 mg/day of prednisolone. American Thoracic Society/European Respiratory Society/World Association of Sarcoidosis and Other Granulomatous Disorders guidelines advise treatment for at least 1 year. Immunosuppressive, cytotoxic, and immunomodulatory agents have been used to treat patients failing or experiencing adverse effects of steroids. Favorable responses have been reported with methotrexate, leflunomide, azathioprine, cyclophosphamide, chlorambucil, cyclosporine A, antimalarials, tumor necrosis factor inhibitors, and thalidomide. Because of potential serious toxicities associated with cyclophosphamide and chlorambucil, these agents are not recommended.9

Our patient presented with pleural sarcoidosis with a pleural effusion and nodules. Treatment with 20 mg of prednisone daily was started initially. Four weeks after discharge, he was still dyspneic and had persistent left pleural effusion. He also had gained a significant amount of weight and developed bilateral lower extremity edema; these were thought to be secondary to prednisone treatment. Steroids were subsequently tapered, and leflunomide was started. His symptoms improved dramatically after 1 month of treatment with leflunomide and steroids, and 3 months later, his pleural effusion had completely resolved.

A 50‐year‐old man presented to the emergency department with progressive shortness of breath for 6 months. He described a dry cough, left‐sided chest pain, malaise, night sweats, and a 15‐pound weight loss. The patient had never smoked cigarettes, but he had been exposed to asbestos and wood dust when working at a sawmill. His physical examination was remarkable for decreased breath sounds at the left lung base. The admission blood tests were within normal limits. Chest radiography and a computed tomography (CT) scan of the chest were performed (the CT scan is shown in Figures 1 and 2). The CT scan showed a left pleural effusion with subpleural and peribronchovascular nodules. Also demonstrated on the CT scan were bilateral hilar and mediastinal lymphadenopathies with faint central calcification. As the left‐sided pleural effusion was initially suspected to be malignant, a thoracentesis was performed, and it revealed an exudative effusion. The total white cell count in fluid was 2100/L (lymphocytes, 76%), and cultures for aerobic and anaerobic bacteria, acid fasting bacilli, and fungi were negative. Cytology was negative for malignant cells. On the basis of the findings in the lung parenchyma and the presence of mediastinal lymphadenopathies, fiber‐optic bronchoscopy with bronchoalveolar lavage, protected specimen brushing, transbronchial needle aspiration, and transbronchial biopsies were performed. Mediastinal lymph node cytology was negative for malignant cells, whereas transbronchial biopsies showed noncaseating granulomas (Figure 3). At that time, our differential diagnoses of noncaseating granulomas included mycobacterium infection (although this usually presents caseating granulomas), berylliosis, histoplasmosis, and sarcoidosis. The tuberculin skin test (purified protein derivative) and serology for human immunodeficiency virus were negative. Bronchoalveolar lavage and cultures of lung tissue biopsies as well as needle aspiration from mediastinal lymph nodes were negative for mycobacterial, fungal, and bacterial organisms. The beryllium lymphocyte proliferation test was normal. Serologic antibodies for Aspergillus, Blastomyces, Coccidioides, and Histoplasma were negative. The urinary Histoplasma antigen was negative as well. The Department of Infectious Diseases was consulted, and an empirical treatment for histoplasmosis with itraconazole was started on the basis of the residence of the patient and the presence of noncaseating granulomas. After 1 month of antifungal treatment, there was no significant improvement. Video‐assisted thoracoscopic surgery with pleural biopsy was performed because of persistent pleural effusion and concern about an underlying infectious or malignant process. Pleural biopsies showed noncaseating granulomas (Figure 4). Pleural fluid was sent for adenosine deaminase (17 U/L) and flow cytometry (CD4/CD8 2.71). Cultures and cytology remained negative. A diagnosis of stage 2 sarcoidosis with pleural involvement was made, and treatment with prednisone was started.

Figure 1

Figure 2

Figure 3

Figure 4

Discussion

The overall prevalence of pleural involvement in sarcoidosis is about 3%. Patients with pleural sarcoidosis tend to be between 30 and 50 years of age, in contrast to the usual presentation of sarcoidosis between 20 and 30 years of age. The most common forms of pleural involvement are pleural effusions, pneumothorax, pleural thickening, and pleural nodules.1 Most effusions are usually small or modest in size, with few reports describing massive effusions.2 Recurrent pleural and pericardial effusions due to sarcoidosis have been reported as well.3 The fluid is typically a lymphocytic exudate, and almost all cases describe a CD4 predominant lymphocytic effusion with CD4/CD8 ratios ranging from 2.35 to 8.6.1 The presence of bloody pleural effusions in sarcoidosis most likely represents the rupture of small vessels that are compressed or infiltrated by granulomas.4

The majority of patients with reported sarcoid pleural effusions have stage 2 disease. With the progression of parenchymal disease, the prevalence of pleural effusions decreases, whereas pleural thickening and pneumothorax increase.5 It is important to emphasize that 40% of pleural effusions in sarcoidosis may be due to other causes, such as tuberculosis and mesothelioma. Our patient was initially treated with itraconazole as histoplasmosis is most prevalent in the Central and Eastern United States, especially in Ohio River valleys, where this patient lived.

The prevalence of a pneumothorax in sarcoidosis is up to 4%.1 Pleural thickening can be demonstrated in 11% to 71% of patients with pleural sarcoidosis, and 10% to 20% of these cases have thickening without effusion. Detection of subpleural nodules and cysts has been possible since the introduction of high‐resolution CT scans. Their prevalence in sarcoidosis ranges from 22% to 76%, and they are often described as masses that correspond to nodules seen in both parietal and visceral surfaces. Hilar or mediastinal lymphadenopathy is present on CT in 47% to 94% of patients with sarcoidosis. Lymph node enlargement is usually bilateral, most commonly with right‐sided predominance. The most involved stations are the right lower paratracheal, right hilar, subcarinal, aortopulmonary window, and right interlobar stations. Nodal calcification is noted in 53% with eggshell calcification present in 9%. The enlargement of internal mammary and pericardial lymph nodes requires the exclusion of lymphoma.6

The management of pleural sarcoidosis should be individualized because a majority of these effusions resolve spontaneously in 1 to 3 months.5 There have been reports of resolution in 2 weeks with steroid therapy. Incomplete resolution of the pleural effusions with progression to chronic pleural thickening or a trapped lung has been reported. There is agreement that oral corticosteroid treatment should be considered in patients with severe persistent or progressively worsening respiratory symptoms or declining lung function. Severe symptoms can be considered as those that interfere with essential aspects of the patient's daily life.7 The initial dosage of oral prednisone recommended by the American Thoracic Society, the European Respiratory Society, and the World Association of Sarcoidosis and Other Granulomatous Disorders guidelines is 20 to 40 mg/day.8 Further evaluation is recommended after 1 to 3 months. If the patient responds, the dose should be reduced gradually to a maintenance dose, such as 5 to 15 mg/day of prednisolone. American Thoracic Society/European Respiratory Society/World Association of Sarcoidosis and Other Granulomatous Disorders guidelines advise treatment for at least 1 year. Immunosuppressive, cytotoxic, and immunomodulatory agents have been used to treat patients failing or experiencing adverse effects of steroids. Favorable responses have been reported with methotrexate, leflunomide, azathioprine, cyclophosphamide, chlorambucil, cyclosporine A, antimalarials, tumor necrosis factor inhibitors, and thalidomide. Because of potential serious toxicities associated with cyclophosphamide and chlorambucil, these agents are not recommended.9

Our patient presented with pleural sarcoidosis with a pleural effusion and nodules. Treatment with 20 mg of prednisone daily was started initially. Four weeks after discharge, he was still dyspneic and had persistent left pleural effusion. He also had gained a significant amount of weight and developed bilateral lower extremity edema; these were thought to be secondary to prednisone treatment. Steroids were subsequently tapered, and leflunomide was started. His symptoms improved dramatically after 1 month of treatment with leflunomide and steroids, and 3 months later, his pleural effusion had completely resolved.

I was 12 years old before I knew her actual name, Le Thi Canh, because we always called her Ba ngoai. She was my grandmother, and this is the story of how she died.

To see my grandmother for the warrior woman that she was to me, you have to know that her farmer father sent her to the city for schooling because he didn't know what else to do with a daughter who was so smart. In early 20th‐century Vietnam, this was an unusual thing to do with a girl. She met my grandfather there, when he was a campus activist, helping him hand out nationalist leaflets. He introduced her to his Communist friends. After the French jailed my grandfather, my grandmother courted him by sending him long letters and care packages while he was in prison. When he was finally released, they married and started a family while he struggled financially as a newspaper publisher and at other odd jobs. But in 1947 his own Communist comrades killed him as part of a party purge. He had been forewarned, and opted to go quietly rather than try to escape because he was promised that this would guarantee her safety. Before they killed him, somewhere in the mountains, my grandfather gave a soldier friend a poem he wrote for his wife. When she told me this story 8 years ago, more than half a century later, she recited his love missive from memory.

At the time of her husband's death, my grandmother had 6 children, the last born just a few weeks before. After a few years of scraping by (she ran her own one‐room school for a while), she decided to leave Hanoi, and migrated south to Saigon with her brood. She was a famously strict parent, to hear my aunts and uncles tell it. She watched them like a hawk, worked full‐time, put them all through school, and eventually rose to a leadership position in the Ministry of Social Work in South Vietnam. My memories of Saigon life are punctuated by scenes of siblings and cousins running around at her regal house, yellow stucco with porticos and black iron gates, at a corner turn in the road, past a cemetery.

On this side of the world, to see her, you would never have thought that my grandmother had led such an epic life. She never worked again after immigrating with us in 1975. She lived on Social Security checks, gardened, said Buddhist prayers, and was nanny to her grandchildren. She watched soap operas religiously, and could report their full plot lines while sitting and knitting. She bundled her many sadnesses in a contented, 4‐foot 9‐inch frame.

Having no home of her own, she would move from one child's house to another every few months so as not to wear out her welcome. But her children lived in Pennsylvania, New Jersey, Florida, Texas, and Maryland. And in most of these cities, she had a different primary care physician. She has 21 grandchildren; 8 of us are physicians. Yet the aunts and uncles told us very little about her medical care. She preferred older‐generation Vietnamese physicians and I'm not sure that they were all competent, but her children did not want to argue with an octogenarian war survivor, and we deferred to their judgment. So we would find out only incidentally, for example, that a doctor prescribed her tuberculosis drugs for a visit to Vietnam.

For many years, Ba ngoai had no major medical problems. She was hypothyroid and hypertensive but on medication and generally high functioning. She had a lumpectomy for early‐stage breast cancer. Then, a year or so after she told me the story of my grandfather's death, Alzheimer's set in. It became harder for her to report symptoms reliably, and she became mildly depressed. Her grandchildren were now birthing our own babies, and we offered these as a distraction, trying to surround her with celebrations of these new fruits of her life labors.

Ba ngoai's decline worsened 3 years ago. She became more easily fatigued, depressed, and confused. A few months before she died, she started to get dyspneic, and couldn't go for short walks any more. In retrospect, I think that her prescription for thyroid replacement somehow fell through the cracks, probably in the transfer of care from one city to another, although there remains a great deal of confusion in the family about exactly what happened. Her thyroxine levels dwindled. One evening in October of that year, at my uncle's house in Maryland, she became severely short of breath and nearly unconscious. They called her Maryland PCP, who sent her to an emergency room at a local hospital. She was admitted in severe congestive heart failure. When the hospitalist spoke with my mother and uncle, he explained that he could take some fluid off her lungs, but that she might need to be intubated and admitted to intensive care. Looking back, I guessed that she probably needed pressors and invasive monitoring. He asked them, Is this what you want? My uncle said, No, it's not. And the hospitalist and the huddle of relatives decided she should come home.

The question was, into whose care would she be discharged? My elders were wary of contacting her PCP, partly because some blamed him for not catching and addressing her symptoms sooner, partly because to even confront him with this perception would cause him, and hence them, to lose face. This seemed too excruciating a scenario to them.

So at last, my uncle called my brother, the oldest grandchild and a very talented clinician. My brother is a pulmonologist, the kind of physician who once did a history and physical on a patient complaining only of Really feeling bad, Doc, and confidently started a steroid infusion before returning the next day with test results confirming his suspicion of Wegener's granulomatosis. He took my grandmother's physical by phone, and told my uncle to increase her furosemide dose. Then he said, I'm on call, but I'll be down there tomorrow. Call everyone together. Most of my relatives were already in town; they had come at news of her decline. She became alert enough for a couple of days to see and recognize most of the faces around her, like so many markers on a long journey. And then she died, slipped away.

I find it hard to define good coordination of care. My instinct as a researcher is to list measurable elements, but the tools we currently have generate metrics that are either reductionistsuch as how rapidly a physician returns a patient's callor so global that they no longer seem actionablesuch as patient satisfaction. But if such metrics set the goal in the distance, it seems useful to also define its counterpartdiscoordinationas a marker of the reality we would like to leave behind us as far as possible. Discoordination includes elements of discontinuity (lost patient history), fragmentation (actions by multiple players), overuse and/or inappropriate use of services, and ultimately, ineffective care (that is, the patient's needs go unmet).

My experience of discoordination was that of a Rube Goldberg contraption. It's composed of innumerable subtasks, each cleverly designed as the easiest solution to a seemingly short‐term problem, as quick fixes, but that in aggregate generate such chaos that the ultimate purpose is lost. They include acts of denial, lies to avoid embarrassment or conflict, and choices of convenience. My mother and her siblings accommodated my grandmother's choice of physicians by (secretly) not always adhering to care recommendations they didn't agree with, instead of challenging her. They took her to different physicians in different cities rather than risk embarrassing (due to an exaggerated sense of the smallness of the Vietnamese community) any one physician by dropping him. Her grandchildren, despite our medical training, found it culturally easier to defer to our elders than to intervene in substandard care. And none of her physicians aggressively followed up to ensure that a frail Alzheimer's patient was getting the care she needed. This is not to suggest that coordination is a simple task because Rube Goldberg machines make simple tasks complicated. Rather, it is a depiction of how indirectly we tend to address the problem.

I imagine a different course of events for my grandmother in the absence of discoordination. What if her children and physicians had understood and acknowledged to one another that her care was fragmented and therefore suboptimal? What if we grandchildren had confronted both Ba Ngoai and our parents sooner about their choice of physicians and offered to take on more of the burden of helping with her care decisions? Would we, as physicians, have been better able to ensure that her providers made rational clinical decisions? And what if she and her family had consistently recognized a single physician as her medical home? Snowbirding is hardly a rare phenomenon among Medicare patients; we could have designated one physician as primarily responsible for coordinating her care even without limiting her travel.

Care coordination is an inherently human activity. Supportive elements such as efficient transfer of medical information, resources for patient education and self‐care, and adequate reimbursement can take us to the brink of, but not actually bridge, the chasm that we want to cross. Traversing that divide sometimes requires settling turf issues over undesirable responsibilities between different physicians and between physicians and other providers; clarifying who has primary responsibility for different types of decisions (I lead on cardiac issues and her son leads on health maintenance); and the violation of cultural norms of patients, families, and/or providers. These can be uncomfortable, unpleasant conversations that at times seem beside the point. But in aggregate, they are the work of coordination, because they force us to align our expectations of one another. No level of information technology could have dismantled the Rube Goldberg machine that trapped my grandmother. Her last of many lessons for me was that emotional courage, honesty, and perseverance offer a much more direct path through the muck.

I was 12 years old before I knew her actual name, Le Thi Canh, because we always called her Ba ngoai. She was my grandmother, and this is the story of how she died.

To see my grandmother for the warrior woman that she was to me, you have to know that her farmer father sent her to the city for schooling because he didn't know what else to do with a daughter who was so smart. In early 20th‐century Vietnam, this was an unusual thing to do with a girl. She met my grandfather there, when he was a campus activist, helping him hand out nationalist leaflets. He introduced her to his Communist friends. After the French jailed my grandfather, my grandmother courted him by sending him long letters and care packages while he was in prison. When he was finally released, they married and started a family while he struggled financially as a newspaper publisher and at other odd jobs. But in 1947 his own Communist comrades killed him as part of a party purge. He had been forewarned, and opted to go quietly rather than try to escape because he was promised that this would guarantee her safety. Before they killed him, somewhere in the mountains, my grandfather gave a soldier friend a poem he wrote for his wife. When she told me this story 8 years ago, more than half a century later, she recited his love missive from memory.

At the time of her husband's death, my grandmother had 6 children, the last born just a few weeks before. After a few years of scraping by (she ran her own one‐room school for a while), she decided to leave Hanoi, and migrated south to Saigon with her brood. She was a famously strict parent, to hear my aunts and uncles tell it. She watched them like a hawk, worked full‐time, put them all through school, and eventually rose to a leadership position in the Ministry of Social Work in South Vietnam. My memories of Saigon life are punctuated by scenes of siblings and cousins running around at her regal house, yellow stucco with porticos and black iron gates, at a corner turn in the road, past a cemetery.

On this side of the world, to see her, you would never have thought that my grandmother had led such an epic life. She never worked again after immigrating with us in 1975. She lived on Social Security checks, gardened, said Buddhist prayers, and was nanny to her grandchildren. She watched soap operas religiously, and could report their full plot lines while sitting and knitting. She bundled her many sadnesses in a contented, 4‐foot 9‐inch frame.

Having no home of her own, she would move from one child's house to another every few months so as not to wear out her welcome. But her children lived in Pennsylvania, New Jersey, Florida, Texas, and Maryland. And in most of these cities, she had a different primary care physician. She has 21 grandchildren; 8 of us are physicians. Yet the aunts and uncles told us very little about her medical care. She preferred older‐generation Vietnamese physicians and I'm not sure that they were all competent, but her children did not want to argue with an octogenarian war survivor, and we deferred to their judgment. So we would find out only incidentally, for example, that a doctor prescribed her tuberculosis drugs for a visit to Vietnam.

For many years, Ba ngoai had no major medical problems. She was hypothyroid and hypertensive but on medication and generally high functioning. She had a lumpectomy for early‐stage breast cancer. Then, a year or so after she told me the story of my grandfather's death, Alzheimer's set in. It became harder for her to report symptoms reliably, and she became mildly depressed. Her grandchildren were now birthing our own babies, and we offered these as a distraction, trying to surround her with celebrations of these new fruits of her life labors.

Ba ngoai's decline worsened 3 years ago. She became more easily fatigued, depressed, and confused. A few months before she died, she started to get dyspneic, and couldn't go for short walks any more. In retrospect, I think that her prescription for thyroid replacement somehow fell through the cracks, probably in the transfer of care from one city to another, although there remains a great deal of confusion in the family about exactly what happened. Her thyroxine levels dwindled. One evening in October of that year, at my uncle's house in Maryland, she became severely short of breath and nearly unconscious. They called her Maryland PCP, who sent her to an emergency room at a local hospital. She was admitted in severe congestive heart failure. When the hospitalist spoke with my mother and uncle, he explained that he could take some fluid off her lungs, but that she might need to be intubated and admitted to intensive care. Looking back, I guessed that she probably needed pressors and invasive monitoring. He asked them, Is this what you want? My uncle said, No, it's not. And the hospitalist and the huddle of relatives decided she should come home.

The question was, into whose care would she be discharged? My elders were wary of contacting her PCP, partly because some blamed him for not catching and addressing her symptoms sooner, partly because to even confront him with this perception would cause him, and hence them, to lose face. This seemed too excruciating a scenario to them.

So at last, my uncle called my brother, the oldest grandchild and a very talented clinician. My brother is a pulmonologist, the kind of physician who once did a history and physical on a patient complaining only of Really feeling bad, Doc, and confidently started a steroid infusion before returning the next day with test results confirming his suspicion of Wegener's granulomatosis. He took my grandmother's physical by phone, and told my uncle to increase her furosemide dose. Then he said, I'm on call, but I'll be down there tomorrow. Call everyone together. Most of my relatives were already in town; they had come at news of her decline. She became alert enough for a couple of days to see and recognize most of the faces around her, like so many markers on a long journey. And then she died, slipped away.

I find it hard to define good coordination of care. My instinct as a researcher is to list measurable elements, but the tools we currently have generate metrics that are either reductionistsuch as how rapidly a physician returns a patient's callor so global that they no longer seem actionablesuch as patient satisfaction. But if such metrics set the goal in the distance, it seems useful to also define its counterpartdiscoordinationas a marker of the reality we would like to leave behind us as far as possible. Discoordination includes elements of discontinuity (lost patient history), fragmentation (actions by multiple players), overuse and/or inappropriate use of services, and ultimately, ineffective care (that is, the patient's needs go unmet).

My experience of discoordination was that of a Rube Goldberg contraption. It's composed of innumerable subtasks, each cleverly designed as the easiest solution to a seemingly short‐term problem, as quick fixes, but that in aggregate generate such chaos that the ultimate purpose is lost. They include acts of denial, lies to avoid embarrassment or conflict, and choices of convenience. My mother and her siblings accommodated my grandmother's choice of physicians by (secretly) not always adhering to care recommendations they didn't agree with, instead of challenging her. They took her to different physicians in different cities rather than risk embarrassing (due to an exaggerated sense of the smallness of the Vietnamese community) any one physician by dropping him. Her grandchildren, despite our medical training, found it culturally easier to defer to our elders than to intervene in substandard care. And none of her physicians aggressively followed up to ensure that a frail Alzheimer's patient was getting the care she needed. This is not to suggest that coordination is a simple task because Rube Goldberg machines make simple tasks complicated. Rather, it is a depiction of how indirectly we tend to address the problem.

I imagine a different course of events for my grandmother in the absence of discoordination. What if her children and physicians had understood and acknowledged to one another that her care was fragmented and therefore suboptimal? What if we grandchildren had confronted both Ba Ngoai and our parents sooner about their choice of physicians and offered to take on more of the burden of helping with her care decisions? Would we, as physicians, have been better able to ensure that her providers made rational clinical decisions? And what if she and her family had consistently recognized a single physician as her medical home? Snowbirding is hardly a rare phenomenon among Medicare patients; we could have designated one physician as primarily responsible for coordinating her care even without limiting her travel.

Care coordination is an inherently human activity. Supportive elements such as efficient transfer of medical information, resources for patient education and self‐care, and adequate reimbursement can take us to the brink of, but not actually bridge, the chasm that we want to cross. Traversing that divide sometimes requires settling turf issues over undesirable responsibilities between different physicians and between physicians and other providers; clarifying who has primary responsibility for different types of decisions (I lead on cardiac issues and her son leads on health maintenance); and the violation of cultural norms of patients, families, and/or providers. These can be uncomfortable, unpleasant conversations that at times seem beside the point. But in aggregate, they are the work of coordination, because they force us to align our expectations of one another. No level of information technology could have dismantled the Rube Goldberg machine that trapped my grandmother. Her last of many lessons for me was that emotional courage, honesty, and perseverance offer a much more direct path through the muck.

I was 12 years old before I knew her actual name, Le Thi Canh, because we always called her Ba ngoai. She was my grandmother, and this is the story of how she died.

To see my grandmother for the warrior woman that she was to me, you have to know that her farmer father sent her to the city for schooling because he didn't know what else to do with a daughter who was so smart. In early 20th‐century Vietnam, this was an unusual thing to do with a girl. She met my grandfather there, when he was a campus activist, helping him hand out nationalist leaflets. He introduced her to his Communist friends. After the French jailed my grandfather, my grandmother courted him by sending him long letters and care packages while he was in prison. When he was finally released, they married and started a family while he struggled financially as a newspaper publisher and at other odd jobs. But in 1947 his own Communist comrades killed him as part of a party purge. He had been forewarned, and opted to go quietly rather than try to escape because he was promised that this would guarantee her safety. Before they killed him, somewhere in the mountains, my grandfather gave a soldier friend a poem he wrote for his wife. When she told me this story 8 years ago, more than half a century later, she recited his love missive from memory.

At the time of her husband's death, my grandmother had 6 children, the last born just a few weeks before. After a few years of scraping by (she ran her own one‐room school for a while), she decided to leave Hanoi, and migrated south to Saigon with her brood. She was a famously strict parent, to hear my aunts and uncles tell it. She watched them like a hawk, worked full‐time, put them all through school, and eventually rose to a leadership position in the Ministry of Social Work in South Vietnam. My memories of Saigon life are punctuated by scenes of siblings and cousins running around at her regal house, yellow stucco with porticos and black iron gates, at a corner turn in the road, past a cemetery.

On this side of the world, to see her, you would never have thought that my grandmother had led such an epic life. She never worked again after immigrating with us in 1975. She lived on Social Security checks, gardened, said Buddhist prayers, and was nanny to her grandchildren. She watched soap operas religiously, and could report their full plot lines while sitting and knitting. She bundled her many sadnesses in a contented, 4‐foot 9‐inch frame.

Having no home of her own, she would move from one child's house to another every few months so as not to wear out her welcome. But her children lived in Pennsylvania, New Jersey, Florida, Texas, and Maryland. And in most of these cities, she had a different primary care physician. She has 21 grandchildren; 8 of us are physicians. Yet the aunts and uncles told us very little about her medical care. She preferred older‐generation Vietnamese physicians and I'm not sure that they were all competent, but her children did not want to argue with an octogenarian war survivor, and we deferred to their judgment. So we would find out only incidentally, for example, that a doctor prescribed her tuberculosis drugs for a visit to Vietnam.

For many years, Ba ngoai had no major medical problems. She was hypothyroid and hypertensive but on medication and generally high functioning. She had a lumpectomy for early‐stage breast cancer. Then, a year or so after she told me the story of my grandfather's death, Alzheimer's set in. It became harder for her to report symptoms reliably, and she became mildly depressed. Her grandchildren were now birthing our own babies, and we offered these as a distraction, trying to surround her with celebrations of these new fruits of her life labors.

Ba ngoai's decline worsened 3 years ago. She became more easily fatigued, depressed, and confused. A few months before she died, she started to get dyspneic, and couldn't go for short walks any more. In retrospect, I think that her prescription for thyroid replacement somehow fell through the cracks, probably in the transfer of care from one city to another, although there remains a great deal of confusion in the family about exactly what happened. Her thyroxine levels dwindled. One evening in October of that year, at my uncle's house in Maryland, she became severely short of breath and nearly unconscious. They called her Maryland PCP, who sent her to an emergency room at a local hospital. She was admitted in severe congestive heart failure. When the hospitalist spoke with my mother and uncle, he explained that he could take some fluid off her lungs, but that she might need to be intubated and admitted to intensive care. Looking back, I guessed that she probably needed pressors and invasive monitoring. He asked them, Is this what you want? My uncle said, No, it's not. And the hospitalist and the huddle of relatives decided she should come home.

The question was, into whose care would she be discharged? My elders were wary of contacting her PCP, partly because some blamed him for not catching and addressing her symptoms sooner, partly because to even confront him with this perception would cause him, and hence them, to lose face. This seemed too excruciating a scenario to them.

So at last, my uncle called my brother, the oldest grandchild and a very talented clinician. My brother is a pulmonologist, the kind of physician who once did a history and physical on a patient complaining only of Really feeling bad, Doc, and confidently started a steroid infusion before returning the next day with test results confirming his suspicion of Wegener's granulomatosis. He took my grandmother's physical by phone, and told my uncle to increase her furosemide dose. Then he said, I'm on call, but I'll be down there tomorrow. Call everyone together. Most of my relatives were already in town; they had come at news of her decline. She became alert enough for a couple of days to see and recognize most of the faces around her, like so many markers on a long journey. And then she died, slipped away.

I find it hard to define good coordination of care. My instinct as a researcher is to list measurable elements, but the tools we currently have generate metrics that are either reductionistsuch as how rapidly a physician returns a patient's callor so global that they no longer seem actionablesuch as patient satisfaction. But if such metrics set the goal in the distance, it seems useful to also define its counterpartdiscoordinationas a marker of the reality we would like to leave behind us as far as possible. Discoordination includes elements of discontinuity (lost patient history), fragmentation (actions by multiple players), overuse and/or inappropriate use of services, and ultimately, ineffective care (that is, the patient's needs go unmet).

My experience of discoordination was that of a Rube Goldberg contraption. It's composed of innumerable subtasks, each cleverly designed as the easiest solution to a seemingly short‐term problem, as quick fixes, but that in aggregate generate such chaos that the ultimate purpose is lost. They include acts of denial, lies to avoid embarrassment or conflict, and choices of convenience. My mother and her siblings accommodated my grandmother's choice of physicians by (secretly) not always adhering to care recommendations they didn't agree with, instead of challenging her. They took her to different physicians in different cities rather than risk embarrassing (due to an exaggerated sense of the smallness of the Vietnamese community) any one physician by dropping him. Her grandchildren, despite our medical training, found it culturally easier to defer to our elders than to intervene in substandard care. And none of her physicians aggressively followed up to ensure that a frail Alzheimer's patient was getting the care she needed. This is not to suggest that coordination is a simple task because Rube Goldberg machines make simple tasks complicated. Rather, it is a depiction of how indirectly we tend to address the problem.

I imagine a different course of events for my grandmother in the absence of discoordination. What if her children and physicians had understood and acknowledged to one another that her care was fragmented and therefore suboptimal? What if we grandchildren had confronted both Ba Ngoai and our parents sooner about their choice of physicians and offered to take on more of the burden of helping with her care decisions? Would we, as physicians, have been better able to ensure that her providers made rational clinical decisions? And what if she and her family had consistently recognized a single physician as her medical home? Snowbirding is hardly a rare phenomenon among Medicare patients; we could have designated one physician as primarily responsible for coordinating her care even without limiting her travel.

Care coordination is an inherently human activity. Supportive elements such as efficient transfer of medical information, resources for patient education and self‐care, and adequate reimbursement can take us to the brink of, but not actually bridge, the chasm that we want to cross. Traversing that divide sometimes requires settling turf issues over undesirable responsibilities between different physicians and between physicians and other providers; clarifying who has primary responsibility for different types of decisions (I lead on cardiac issues and her son leads on health maintenance); and the violation of cultural norms of patients, families, and/or providers. These can be uncomfortable, unpleasant conversations that at times seem beside the point. But in aggregate, they are the work of coordination, because they force us to align our expectations of one another. No level of information technology could have dismantled the Rube Goldberg machine that trapped my grandmother. Her last of many lessons for me was that emotional courage, honesty, and perseverance offer a much more direct path through the muck.

Hospitalists recognize the importance of the care transition from the inpatient setting to the outpatient setting, despite being described as causing a divorce between inpatient and outpatient care.1 If you do not believe this, just glance at the table of contents for this issue of the Journal of Hospital Medicine, which has 5 reports on research about various aspects of the hospital discharge transition complemented by an eloquent story of how a hospitalist facilitated the care coordination of one family's matriarch.2 An accompanying editorial proposes that hospitalists embrace the need of patients and their caregivers for care coordination.3 Thankfully, a growing number of academic hospitalists are focusing their efforts on identifying problems in the process and evaluating potential interventions to optimize it.

The hospital discharge process commonly has been an afterthought, concluding a typically intense experience for patients, some of whom may have begun the episode of hospitalization near death. After diagnostic evaluations and treatments, a patient has achieved stable enough status to be discharged home, and the inpatient physician has signed off with a simple may go in the written orders. The physician may feel absolved of responsibility as he expects the nurses to take care of instructions and to find transportation home for the patient. Unfortunately, this experience often is consistent with Webster's definition of discharge: to relieve of a charge, load, or burden unload release from an obligation. Some patients may feel like a Nolan Ryan fastball flying out of the hospital, but with no one to catch them.

Recognizing how the hospital discharge transition to home can be a perilous process fraught with failure,4 we laid out a research agenda for transitions of care. We are gratified to see the robust response from researchers published in this issue of the Journal of Hospital Medicine. The studies range from the description of a new tool to assess patients' mobility before discharge5 to evidence that the length of stay is prolonged (ie, delayed discharge) when the discharge diagnosis differs from that made on admission.6 Chen and colleagues analyzed the timing of discharge during the day and found that the duration of the discharge process was influenced by the need for consultation or a procedure prior to discharge; this finding is not surprising to practicing hospitalists. We agree with their conclusion that broad institutional efforts will be needed to facilitate the process. Hospitalists are part of a system and must engage the entire team to improve efficiency.

O'Leary and fellow hospitalists7 at Northwestern Memorial Hospital focused on creating a better discharge summary within their electronic health record with the aim of improved overall quality of the summaries and, just as important, timely completion. Despite some research indicating that absence of adequate communication between primary care providers and inpatient medical teams is not associated with adverse clinical outcomes,8 other research has demonstrated that it does affect outcomes and probably affects rehospitalization rates.9, 10 Moreover, another article in this issue describes a project undertaken at Baylor Health Care System (Dallas, TX) that demonstrated a reduction in emergency department visits and readmissions within 30 days post‐discharge among high‐risk elderly medical patients when a targeted care bundle was used.11 The results from this intervention, which consisted of medication counseling/reconciliation by a clinical pharmacist, condition‐specific enhanced discharge planning by a care coordinator, and phone follow‐up, confirm recent results from 2 similar studies.12, 13 These studies provide support for the idea that straightforward changes in the discharge process can improve patient outcomes.

Today in the United States, hospitalists likely care for the majority of hospitalized older patients.14 We strongly encourage them to use evidence‐based approaches to optimize the discharge process in their hospitals, and fortunately, clear guidance is available. Because of generous funding from the John A. Hartford Foundation, Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) is mentoring 30 hospitals in an effort to implement the BOOST toolkit and improve their discharge transition processes.15 Another cost‐effective method involves the use of transition coaches to help the most vulnerable older patients with complex care needs.16 This approach is now being implemented by more than 100 healthcare organizations worldwide.17

Heartened by these exciting initiatives, we applaud the Society of Hospital Medicine's collaboration with the American College of Physicians, the Society of General Internal Medicine, the American Geriatrics Society, and the Society of Academic Emergency Medicine to produce a consensus policy statement on transitions of care that provides guiding principles for transitions both into and out of the hospital.18 Soon, all hospitalized patients and their caregivers may receive robust education prior to discharge, confirmation of their understanding with the teach‐back approach, medication reconciliation, and clear instructions for follow‐up, and the patient's primary care provider will be aware of all that has happened. Patients should expect nothing less than hospitalists ensuring their seamless transition from hospital to home.

Hospitalists recognize the importance of the care transition from the inpatient setting to the outpatient setting, despite being described as causing a divorce between inpatient and outpatient care.1 If you do not believe this, just glance at the table of contents for this issue of the Journal of Hospital Medicine, which has 5 reports on research about various aspects of the hospital discharge transition complemented by an eloquent story of how a hospitalist facilitated the care coordination of one family's matriarch.2 An accompanying editorial proposes that hospitalists embrace the need of patients and their caregivers for care coordination.3 Thankfully, a growing number of academic hospitalists are focusing their efforts on identifying problems in the process and evaluating potential interventions to optimize it.

The hospital discharge process commonly has been an afterthought, concluding a typically intense experience for patients, some of whom may have begun the episode of hospitalization near death. After diagnostic evaluations and treatments, a patient has achieved stable enough status to be discharged home, and the inpatient physician has signed off with a simple may go in the written orders. The physician may feel absolved of responsibility as he expects the nurses to take care of instructions and to find transportation home for the patient. Unfortunately, this experience often is consistent with Webster's definition of discharge: to relieve of a charge, load, or burden unload release from an obligation. Some patients may feel like a Nolan Ryan fastball flying out of the hospital, but with no one to catch them.

Recognizing how the hospital discharge transition to home can be a perilous process fraught with failure,4 we laid out a research agenda for transitions of care. We are gratified to see the robust response from researchers published in this issue of the Journal of Hospital Medicine. The studies range from the description of a new tool to assess patients' mobility before discharge5 to evidence that the length of stay is prolonged (ie, delayed discharge) when the discharge diagnosis differs from that made on admission.6 Chen and colleagues analyzed the timing of discharge during the day and found that the duration of the discharge process was influenced by the need for consultation or a procedure prior to discharge; this finding is not surprising to practicing hospitalists. We agree with their conclusion that broad institutional efforts will be needed to facilitate the process. Hospitalists are part of a system and must engage the entire team to improve efficiency.

O'Leary and fellow hospitalists7 at Northwestern Memorial Hospital focused on creating a better discharge summary within their electronic health record with the aim of improved overall quality of the summaries and, just as important, timely completion. Despite some research indicating that absence of adequate communication between primary care providers and inpatient medical teams is not associated with adverse clinical outcomes,8 other research has demonstrated that it does affect outcomes and probably affects rehospitalization rates.9, 10 Moreover, another article in this issue describes a project undertaken at Baylor Health Care System (Dallas, TX) that demonstrated a reduction in emergency department visits and readmissions within 30 days post‐discharge among high‐risk elderly medical patients when a targeted care bundle was used.11 The results from this intervention, which consisted of medication counseling/reconciliation by a clinical pharmacist, condition‐specific enhanced discharge planning by a care coordinator, and phone follow‐up, confirm recent results from 2 similar studies.12, 13 These studies provide support for the idea that straightforward changes in the discharge process can improve patient outcomes.

Today in the United States, hospitalists likely care for the majority of hospitalized older patients.14 We strongly encourage them to use evidence‐based approaches to optimize the discharge process in their hospitals, and fortunately, clear guidance is available. Because of generous funding from the John A. Hartford Foundation, Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) is mentoring 30 hospitals in an effort to implement the BOOST toolkit and improve their discharge transition processes.15 Another cost‐effective method involves the use of transition coaches to help the most vulnerable older patients with complex care needs.16 This approach is now being implemented by more than 100 healthcare organizations worldwide.17

Heartened by these exciting initiatives, we applaud the Society of Hospital Medicine's collaboration with the American College of Physicians, the Society of General Internal Medicine, the American Geriatrics Society, and the Society of Academic Emergency Medicine to produce a consensus policy statement on transitions of care that provides guiding principles for transitions both into and out of the hospital.18 Soon, all hospitalized patients and their caregivers may receive robust education prior to discharge, confirmation of their understanding with the teach‐back approach, medication reconciliation, and clear instructions for follow‐up, and the patient's primary care provider will be aware of all that has happened. Patients should expect nothing less than hospitalists ensuring their seamless transition from hospital to home.

Hospitalists recognize the importance of the care transition from the inpatient setting to the outpatient setting, despite being described as causing a divorce between inpatient and outpatient care.1 If you do not believe this, just glance at the table of contents for this issue of the Journal of Hospital Medicine, which has 5 reports on research about various aspects of the hospital discharge transition complemented by an eloquent story of how a hospitalist facilitated the care coordination of one family's matriarch.2 An accompanying editorial proposes that hospitalists embrace the need of patients and their caregivers for care coordination.3 Thankfully, a growing number of academic hospitalists are focusing their efforts on identifying problems in the process and evaluating potential interventions to optimize it.

The hospital discharge process commonly has been an afterthought, concluding a typically intense experience for patients, some of whom may have begun the episode of hospitalization near death. After diagnostic evaluations and treatments, a patient has achieved stable enough status to be discharged home, and the inpatient physician has signed off with a simple may go in the written orders. The physician may feel absolved of responsibility as he expects the nurses to take care of instructions and to find transportation home for the patient. Unfortunately, this experience often is consistent with Webster's definition of discharge: to relieve of a charge, load, or burden unload release from an obligation. Some patients may feel like a Nolan Ryan fastball flying out of the hospital, but with no one to catch them.

Recognizing how the hospital discharge transition to home can be a perilous process fraught with failure,4 we laid out a research agenda for transitions of care. We are gratified to see the robust response from researchers published in this issue of the Journal of Hospital Medicine. The studies range from the description of a new tool to assess patients' mobility before discharge5 to evidence that the length of stay is prolonged (ie, delayed discharge) when the discharge diagnosis differs from that made on admission.6 Chen and colleagues analyzed the timing of discharge during the day and found that the duration of the discharge process was influenced by the need for consultation or a procedure prior to discharge; this finding is not surprising to practicing hospitalists. We agree with their conclusion that broad institutional efforts will be needed to facilitate the process. Hospitalists are part of a system and must engage the entire team to improve efficiency.

O'Leary and fellow hospitalists7 at Northwestern Memorial Hospital focused on creating a better discharge summary within their electronic health record with the aim of improved overall quality of the summaries and, just as important, timely completion. Despite some research indicating that absence of adequate communication between primary care providers and inpatient medical teams is not associated with adverse clinical outcomes,8 other research has demonstrated that it does affect outcomes and probably affects rehospitalization rates.9, 10 Moreover, another article in this issue describes a project undertaken at Baylor Health Care System (Dallas, TX) that demonstrated a reduction in emergency department visits and readmissions within 30 days post‐discharge among high‐risk elderly medical patients when a targeted care bundle was used.11 The results from this intervention, which consisted of medication counseling/reconciliation by a clinical pharmacist, condition‐specific enhanced discharge planning by a care coordinator, and phone follow‐up, confirm recent results from 2 similar studies.12, 13 These studies provide support for the idea that straightforward changes in the discharge process can improve patient outcomes.

Today in the United States, hospitalists likely care for the majority of hospitalized older patients.14 We strongly encourage them to use evidence‐based approaches to optimize the discharge process in their hospitals, and fortunately, clear guidance is available. Because of generous funding from the John A. Hartford Foundation, Project BOOST (Better Outcomes for Older Adults Through Safe Transitions) is mentoring 30 hospitals in an effort to implement the BOOST toolkit and improve their discharge transition processes.15 Another cost‐effective method involves the use of transition coaches to help the most vulnerable older patients with complex care needs.16 This approach is now being implemented by more than 100 healthcare organizations worldwide.17

Heartened by these exciting initiatives, we applaud the Society of Hospital Medicine's collaboration with the American College of Physicians, the Society of General Internal Medicine, the American Geriatrics Society, and the Society of Academic Emergency Medicine to produce a consensus policy statement on transitions of care that provides guiding principles for transitions both into and out of the hospital.18 Soon, all hospitalized patients and their caregivers may receive robust education prior to discharge, confirmation of their understanding with the teach‐back approach, medication reconciliation, and clear instructions for follow‐up, and the patient's primary care provider will be aware of all that has happened. Patients should expect nothing less than hospitalists ensuring their seamless transition from hospital to home.

Transient ischemic attacks (TIAs) are common and represent a clarion call to action to prevent disabling stroke. Incidence estimates for TIA range from 37 to 107 per 100,000 persons each year.1 Extrapolating from these data, there are likely greater than 100,000 to 300,000 TIAs in the US annually. Within 3 months, approximately 10% of these patients will suffer a stroke, with approximately one‐half of these events occurring within the first 48 hours after the sentinel TIA.26 Nearly two‐thirds of secondary strokes result in disability and 21% are fatal.3 Hospitalists are frequently called to provide care for patients with TIA and, as such, in order to establish an appropriate care plan, they require tools to better predict the likelihood and timing of a disabling stroke.7 In this review we examine the rationale for early aggressive TIA evaluation and treatment in the hospital, overview risk stratification models to identify the patients at highest risk for early recurrent ischemia, and explore application of these tools to admission policy and individualized patient care planning.

Definition

TIA is defined as a brief episode of neurological dysfunction caused by focal brain or retinal ischemia with clinical symptoms typically lasting less than 1 hour and without evidence of brain infarction.8, 9 Prior arbitrary time limits are being abandoned as advanced imaging techniques demonstrate that clinical examination lacks the sensitivity to detect small cerebral infarctions leading to misclassification of as many as 30% to 40% of strokes as TIAs.811 For cases in which imaging is not available, the diagnosis of clinically probable TIA is suggested. Patients with imaging consistent with stroke appear to be at 4‐fold to 10‐fold higher risk for subsequent ischemic events, thus the presence of subclinical infarcts may have clinical importance.2, 12 The majority of TIAs resolve within 1 hour of onset and neurologic deficit continuance beyond this time frame is more consistent with a stroke.13 Continuing symptoms after 1 hour mandates aggressive therapy in lieu of withholding intervention in the hopes of a spontaneous recovery.

Rationale for Hospitalization

Urgent evaluation and treatment within 24 to 48 hours of a TIA is recommended by the National Stroke Association (Table 1).14 These guidelines also recommend hospital admission for high‐risk patients. There are a number of compelling arguments for the hospitalization of a patient at high risk for subsequent stroke.

First, hospitalization offers potential for reduced time to thrombolysis for those patients who have a second ischemic event in the early period following TIA. Outpatients with new ischemic stroke may see hours pass between symptom onset and presentation to the emergency department (ED). This delay frequently places them outside of the thrombolytic window.1517 Hospitalization, assuming a well‐designed inpatient stroke care system, has great potential to reduce this delay. Approximately 50% of the stroke risk following a TIA is evident within 48 hours and rapid thrombolysis, available in an inpatient setting, is associated with improved outcome after stroke.3, 18 A cost‐utility analysis found that a 24‐hour admission for TIA patients to allow tissue plasminogen activator (t‐PA) for recurrent ischemia has a cost‐effectiveness ratio of $55,044 per quality‐adjusted life year with increasing cost effectiveness for the highest risk patients, such as those with a 24‐hour stroke risk of >5%.19

Second, hospital admission often facilitates the reliable and efficient evaluation for etiology and early initiation of secondary prevention. Neuroimaging, carotid ultrasound, echocardiography, and telemetry can be expedited with rapid initiation of proven secondary preventive therapies such as statin treatment, blood pressure control, and antithrombotic therapy. When indicated, carotid revascularization is recommended as soon as possible following TIA, with retrospective reviews suggesting improved outcomes when performed within 2 weeks of the event.1420 In one analysis, a negative association between hospitalization for TIA and subsequent stroke was discovered by review of Canadian population‐based administrative databases.5 While the mechanism for the negative association could not be established, the literature provides some support for hospitalization being associated with decreased risk for second strokes (hazard ratio [HR], 0.73; 95% confidence interval [CI], 0.570.95).5

Theoretically, much of this evaluation and treatment could occur in the outpatient setting but delays commonly seen in outpatient evaluation and the high potential for early second strokes for some patients may make this a risky care plan. Despite the high likelihood for serious outcomes following TIA and clear guidelines for early evaluation and management, current care often lacks a sense of urgency. A 2004 Canadian study revealed that three‐quarters of patients with a TIA were discharged directly from the ED with a resultant delay in diagnostic investigation.4 Over one‐third of patients were discharged without a prescription for antithrombotic therapy. American primary care practice patterns reveal even more significant delays in therapy, with only 2% of patients admitted to a hospital on the day of presentation for TIA, despite 80% of patients presenting for evaluation on the day of symptom onset.21 In this study less than one‐half of patients with atrial fibrillation were started on immediate anticoagulation.21 Further, as many as one‐third of patients did not have any evaluation in the month after the index event.21 Hospitalization for high‐risk patients has the potential to avoid these delays in outpatient evaluation and initiation of therapy.

Still, not all patients will require admission to a hospital setting. American EDs admit approximately one‐half of all TIAs, with regional variability not explained exclusively by clinical characteristics.22 Focusing on identifying the cohort of patients who would most benefit from hospitalization is paramount. In general, hospitalization should be reserved for patients with higher risk of an early secondary stroke. Specifically, admission is generally recommended for patients with crescendo symptoms, TIA on antithrombotic therapy, or symptoms lasting >1 hour.14 Additionally, patients with symptomatic carotid stenosis of 50% and presumed cardioembolic or hypercoagulable etiology merit hospital admission.14 In many cases these etiologies may not be known at time of presentation. Evaluation, such as carotid ultrasound, may not be readily available in the ED to inform the admission decision. Several new scoring systems that utilize routine clinical features available within an hour of presentation have been developed to more objectively assess the risk of secondary stroke following a TIA. The use of these prognostic scoring systems is recommended by the National Stroke Association to aid in triaging this cohort of patients.14

Prognostic Scoring Systems

New Models of Care: An Opportunity for Hospitalists

The key to improving TIA outcomes appears to be more contingent on the speed of evaluation and initiation of appropriate therapy than on the location of the care. The EXPRESS trial studied the effect of an immediate access neurovascular clinic providing urgent evaluation and immediate treatment of nonhospitalized TIA patients versus usual care. Statistically significant reductions were seen in time to evaluation, first treatment prescription, and in 90‐day risk of recurrent stroke (10.3% versus 2.1%, P < 0.0001) after the clinic was changed to the rapid evaluation and treatment model.33

The SOS‐TIA study used a 24‐hour access hospital‐based TIA clinic to evaluate the effects of rapid assessment and interventions on hospital length of stay and clinical outcomes.34 The 90‐day stroke rate was 1.24% (95% CI, 0.712.12), which represents a 79% reduction compared to the predicted stroke rate from the ABCD² scores. With expedited evaluation and treatment, 74% of patients were able to be sent home on the same day.

The results of these 2 new studies provide compelling evidence that rapid evaluation and treatment in the first 48 hours after TIA has the potential to alter outcomes. Unfortunately not all communities have access to same day TIA clinics. Still, these findings should embolden hospitalists to advocate for urgent evaluation, such as neurology and cardiac imaging and carotid evaluation, with immediate initiation of secondary preventive therapy and early surgical intervention when appropriate. In most cases these changes will require process transformations that present prime opportunities for hospitalists to reengineer systems of care.

Incorporating Prognostic Scores into Clinical Practice

Applying the evidence to practice requires calculation of the early risk but also awareness of the community resources available. High‐risk patients with an ABCD² score of 6 or 7 have a very high 8.1% risk of stroke within the next 48 hours. Given the catastrophic outcomes frequently seen after second strokes, these patients warrant inpatient admission to facilitate the immediate initiation of appropriate secondary prevention and potentially shorten time to thrombolysis if an early stroke occurs. Intermediate‐risk patients with ABCD² scores of 4 and 5 have a 4.1% 2‐day risk of stroke and may be considered for admission, hospital observation, or expedited clinic evaluation contingent on local availability. As many as one‐third of TIA patients will be categorized as low risk with a score of 0 to 3. These patients have a 2‐day risk of stroke of only 1% and are likely safe for prompt outpatient evaluation and management. The new, validated, ABCD² score is not a substitute for individualized judgment, but is helpful in developing admission guidelines in cooperation between neurologists, emergency room physicians, and hospitalists, and in using evidence‐based medicine to provide optimal care for the patient presenting with a TIA.

Stroke and TIA arise from identical etiologies, respond to the same secondary preventive measures, and should be considered part of the spectrum of an ischemic cerebral syndrome. Recognizing TIA as a medical emergency with high rates of secondary stroke and subsequent disability allows institution of therapies with appropriate urgency. Hospitalization offers the ability to rapidly coordinate the testing and secondary prevention measures but also, for high‐risk patients, offers the opportunity to reduce the time to thrombolysis for early recurrent strokes. New, validated scoring systems such as the ABCD² score help the hospitalist to decide which patients are appropriate for admission and which can be managed in progressive and traditional outpatient settings.

Transient ischemic attacks (TIAs) are common and represent a clarion call to action to prevent disabling stroke. Incidence estimates for TIA range from 37 to 107 per 100,000 persons each year.1 Extrapolating from these data, there are likely greater than 100,000 to 300,000 TIAs in the US annually. Within 3 months, approximately 10% of these patients will suffer a stroke, with approximately one‐half of these events occurring within the first 48 hours after the sentinel TIA.26 Nearly two‐thirds of secondary strokes result in disability and 21% are fatal.3 Hospitalists are frequently called to provide care for patients with TIA and, as such, in order to establish an appropriate care plan, they require tools to better predict the likelihood and timing of a disabling stroke.7 In this review we examine the rationale for early aggressive TIA evaluation and treatment in the hospital, overview risk stratification models to identify the patients at highest risk for early recurrent ischemia, and explore application of these tools to admission policy and individualized patient care planning.

Definition

TIA is defined as a brief episode of neurological dysfunction caused by focal brain or retinal ischemia with clinical symptoms typically lasting less than 1 hour and without evidence of brain infarction.8, 9 Prior arbitrary time limits are being abandoned as advanced imaging techniques demonstrate that clinical examination lacks the sensitivity to detect small cerebral infarctions leading to misclassification of as many as 30% to 40% of strokes as TIAs.811 For cases in which imaging is not available, the diagnosis of clinically probable TIA is suggested. Patients with imaging consistent with stroke appear to be at 4‐fold to 10‐fold higher risk for subsequent ischemic events, thus the presence of subclinical infarcts may have clinical importance.2, 12 The majority of TIAs resolve within 1 hour of onset and neurologic deficit continuance beyond this time frame is more consistent with a stroke.13 Continuing symptoms after 1 hour mandates aggressive therapy in lieu of withholding intervention in the hopes of a spontaneous recovery.

Rationale for Hospitalization

Urgent evaluation and treatment within 24 to 48 hours of a TIA is recommended by the National Stroke Association (Table 1).14 These guidelines also recommend hospital admission for high‐risk patients. There are a number of compelling arguments for the hospitalization of a patient at high risk for subsequent stroke.

First, hospitalization offers potential for reduced time to thrombolysis for those patients who have a second ischemic event in the early period following TIA. Outpatients with new ischemic stroke may see hours pass between symptom onset and presentation to the emergency department (ED). This delay frequently places them outside of the thrombolytic window.1517 Hospitalization, assuming a well‐designed inpatient stroke care system, has great potential to reduce this delay. Approximately 50% of the stroke risk following a TIA is evident within 48 hours and rapid thrombolysis, available in an inpatient setting, is associated with improved outcome after stroke.3, 18 A cost‐utility analysis found that a 24‐hour admission for TIA patients to allow tissue plasminogen activator (t‐PA) for recurrent ischemia has a cost‐effectiveness ratio of $55,044 per quality‐adjusted life year with increasing cost effectiveness for the highest risk patients, such as those with a 24‐hour stroke risk of >5%.19

Second, hospital admission often facilitates the reliable and efficient evaluation for etiology and early initiation of secondary prevention. Neuroimaging, carotid ultrasound, echocardiography, and telemetry can be expedited with rapid initiation of proven secondary preventive therapies such as statin treatment, blood pressure control, and antithrombotic therapy. When indicated, carotid revascularization is recommended as soon as possible following TIA, with retrospective reviews suggesting improved outcomes when performed within 2 weeks of the event.1420 In one analysis, a negative association between hospitalization for TIA and subsequent stroke was discovered by review of Canadian population‐based administrative databases.5 While the mechanism for the negative association could not be established, the literature provides some support for hospitalization being associated with decreased risk for second strokes (hazard ratio [HR], 0.73; 95% confidence interval [CI], 0.570.95).5

Theoretically, much of this evaluation and treatment could occur in the outpatient setting but delays commonly seen in outpatient evaluation and the high potential for early second strokes for some patients may make this a risky care plan. Despite the high likelihood for serious outcomes following TIA and clear guidelines for early evaluation and management, current care often lacks a sense of urgency. A 2004 Canadian study revealed that three‐quarters of patients with a TIA were discharged directly from the ED with a resultant delay in diagnostic investigation.4 Over one‐third of patients were discharged without a prescription for antithrombotic therapy. American primary care practice patterns reveal even more significant delays in therapy, with only 2% of patients admitted to a hospital on the day of presentation for TIA, despite 80% of patients presenting for evaluation on the day of symptom onset.21 In this study less than one‐half of patients with atrial fibrillation were started on immediate anticoagulation.21 Further, as many as one‐third of patients did not have any evaluation in the month after the index event.21 Hospitalization for high‐risk patients has the potential to avoid these delays in outpatient evaluation and initiation of therapy.

Still, not all patients will require admission to a hospital setting. American EDs admit approximately one‐half of all TIAs, with regional variability not explained exclusively by clinical characteristics.22 Focusing on identifying the cohort of patients who would most benefit from hospitalization is paramount. In general, hospitalization should be reserved for patients with higher risk of an early secondary stroke. Specifically, admission is generally recommended for patients with crescendo symptoms, TIA on antithrombotic therapy, or symptoms lasting >1 hour.14 Additionally, patients with symptomatic carotid stenosis of 50% and presumed cardioembolic or hypercoagulable etiology merit hospital admission.14 In many cases these etiologies may not be known at time of presentation. Evaluation, such as carotid ultrasound, may not be readily available in the ED to inform the admission decision. Several new scoring systems that utilize routine clinical features available within an hour of presentation have been developed to more objectively assess the risk of secondary stroke following a TIA. The use of these prognostic scoring systems is recommended by the National Stroke Association to aid in triaging this cohort of patients.14

Prognostic Scoring Systems

New Models of Care: An Opportunity for Hospitalists

The key to improving TIA outcomes appears to be more contingent on the speed of evaluation and initiation of appropriate therapy than on the location of the care. The EXPRESS trial studied the effect of an immediate access neurovascular clinic providing urgent evaluation and immediate treatment of nonhospitalized TIA patients versus usual care. Statistically significant reductions were seen in time to evaluation, first treatment prescription, and in 90‐day risk of recurrent stroke (10.3% versus 2.1%, P < 0.0001) after the clinic was changed to the rapid evaluation and treatment model.33

The SOS‐TIA study used a 24‐hour access hospital‐based TIA clinic to evaluate the effects of rapid assessment and interventions on hospital length of stay and clinical outcomes.34 The 90‐day stroke rate was 1.24% (95% CI, 0.712.12), which represents a 79% reduction compared to the predicted stroke rate from the ABCD² scores. With expedited evaluation and treatment, 74% of patients were able to be sent home on the same day.

The results of these 2 new studies provide compelling evidence that rapid evaluation and treatment in the first 48 hours after TIA has the potential to alter outcomes. Unfortunately not all communities have access to same day TIA clinics. Still, these findings should embolden hospitalists to advocate for urgent evaluation, such as neurology and cardiac imaging and carotid evaluation, with immediate initiation of secondary preventive therapy and early surgical intervention when appropriate. In most cases these changes will require process transformations that present prime opportunities for hospitalists to reengineer systems of care.

Incorporating Prognostic Scores into Clinical Practice

Applying the evidence to practice requires calculation of the early risk but also awareness of the community resources available. High‐risk patients with an ABCD² score of 6 or 7 have a very high 8.1% risk of stroke within the next 48 hours. Given the catastrophic outcomes frequently seen after second strokes, these patients warrant inpatient admission to facilitate the immediate initiation of appropriate secondary prevention and potentially shorten time to thrombolysis if an early stroke occurs. Intermediate‐risk patients with ABCD² scores of 4 and 5 have a 4.1% 2‐day risk of stroke and may be considered for admission, hospital observation, or expedited clinic evaluation contingent on local availability. As many as one‐third of TIA patients will be categorized as low risk with a score of 0 to 3. These patients have a 2‐day risk of stroke of only 1% and are likely safe for prompt outpatient evaluation and management. The new, validated, ABCD² score is not a substitute for individualized judgment, but is helpful in developing admission guidelines in cooperation between neurologists, emergency room physicians, and hospitalists, and in using evidence‐based medicine to provide optimal care for the patient presenting with a TIA.

Stroke and TIA arise from identical etiologies, respond to the same secondary preventive measures, and should be considered part of the spectrum of an ischemic cerebral syndrome. Recognizing TIA as a medical emergency with high rates of secondary stroke and subsequent disability allows institution of therapies with appropriate urgency. Hospitalization offers the ability to rapidly coordinate the testing and secondary prevention measures but also, for high‐risk patients, offers the opportunity to reduce the time to thrombolysis for early recurrent strokes. New, validated scoring systems such as the ABCD² score help the hospitalist to decide which patients are appropriate for admission and which can be managed in progressive and traditional outpatient settings.

Transient ischemic attacks (TIAs) are common and represent a clarion call to action to prevent disabling stroke. Incidence estimates for TIA range from 37 to 107 per 100,000 persons each year.1 Extrapolating from these data, there are likely greater than 100,000 to 300,000 TIAs in the US annually. Within 3 months, approximately 10% of these patients will suffer a stroke, with approximately one‐half of these events occurring within the first 48 hours after the sentinel TIA.26 Nearly two‐thirds of secondary strokes result in disability and 21% are fatal.3 Hospitalists are frequently called to provide care for patients with TIA and, as such, in order to establish an appropriate care plan, they require tools to better predict the likelihood and timing of a disabling stroke.7 In this review we examine the rationale for early aggressive TIA evaluation and treatment in the hospital, overview risk stratification models to identify the patients at highest risk for early recurrent ischemia, and explore application of these tools to admission policy and individualized patient care planning.

Definition

TIA is defined as a brief episode of neurological dysfunction caused by focal brain or retinal ischemia with clinical symptoms typically lasting less than 1 hour and without evidence of brain infarction.8, 9 Prior arbitrary time limits are being abandoned as advanced imaging techniques demonstrate that clinical examination lacks the sensitivity to detect small cerebral infarctions leading to misclassification of as many as 30% to 40% of strokes as TIAs.811 For cases in which imaging is not available, the diagnosis of clinically probable TIA is suggested. Patients with imaging consistent with stroke appear to be at 4‐fold to 10‐fold higher risk for subsequent ischemic events, thus the presence of subclinical infarcts may have clinical importance.2, 12 The majority of TIAs resolve within 1 hour of onset and neurologic deficit continuance beyond this time frame is more consistent with a stroke.13 Continuing symptoms after 1 hour mandates aggressive therapy in lieu of withholding intervention in the hopes of a spontaneous recovery.

Rationale for Hospitalization

Urgent evaluation and treatment within 24 to 48 hours of a TIA is recommended by the National Stroke Association (Table 1).14 These guidelines also recommend hospital admission for high‐risk patients. There are a number of compelling arguments for the hospitalization of a patient at high risk for subsequent stroke.

First, hospitalization offers potential for reduced time to thrombolysis for those patients who have a second ischemic event in the early period following TIA. Outpatients with new ischemic stroke may see hours pass between symptom onset and presentation to the emergency department (ED). This delay frequently places them outside of the thrombolytic window.1517 Hospitalization, assuming a well‐designed inpatient stroke care system, has great potential to reduce this delay. Approximately 50% of the stroke risk following a TIA is evident within 48 hours and rapid thrombolysis, available in an inpatient setting, is associated with improved outcome after stroke.3, 18 A cost‐utility analysis found that a 24‐hour admission for TIA patients to allow tissue plasminogen activator (t‐PA) for recurrent ischemia has a cost‐effectiveness ratio of $55,044 per quality‐adjusted life year with increasing cost effectiveness for the highest risk patients, such as those with a 24‐hour stroke risk of >5%.19

Second, hospital admission often facilitates the reliable and efficient evaluation for etiology and early initiation of secondary prevention. Neuroimaging, carotid ultrasound, echocardiography, and telemetry can be expedited with rapid initiation of proven secondary preventive therapies such as statin treatment, blood pressure control, and antithrombotic therapy. When indicated, carotid revascularization is recommended as soon as possible following TIA, with retrospective reviews suggesting improved outcomes when performed within 2 weeks of the event.1420 In one analysis, a negative association between hospitalization for TIA and subsequent stroke was discovered by review of Canadian population‐based administrative databases.5 While the mechanism for the negative association could not be established, the literature provides some support for hospitalization being associated with decreased risk for second strokes (hazard ratio [HR], 0.73; 95% confidence interval [CI], 0.570.95).5

Theoretically, much of this evaluation and treatment could occur in the outpatient setting but delays commonly seen in outpatient evaluation and the high potential for early second strokes for some patients may make this a risky care plan. Despite the high likelihood for serious outcomes following TIA and clear guidelines for early evaluation and management, current care often lacks a sense of urgency. A 2004 Canadian study revealed that three‐quarters of patients with a TIA were discharged directly from the ED with a resultant delay in diagnostic investigation.4 Over one‐third of patients were discharged without a prescription for antithrombotic therapy. American primary care practice patterns reveal even more significant delays in therapy, with only 2% of patients admitted to a hospital on the day of presentation for TIA, despite 80% of patients presenting for evaluation on the day of symptom onset.21 In this study less than one‐half of patients with atrial fibrillation were started on immediate anticoagulation.21 Further, as many as one‐third of patients did not have any evaluation in the month after the index event.21 Hospitalization for high‐risk patients has the potential to avoid these delays in outpatient evaluation and initiation of therapy.

Still, not all patients will require admission to a hospital setting. American EDs admit approximately one‐half of all TIAs, with regional variability not explained exclusively by clinical characteristics.22 Focusing on identifying the cohort of patients who would most benefit from hospitalization is paramount. In general, hospitalization should be reserved for patients with higher risk of an early secondary stroke. Specifically, admission is generally recommended for patients with crescendo symptoms, TIA on antithrombotic therapy, or symptoms lasting >1 hour.14 Additionally, patients with symptomatic carotid stenosis of 50% and presumed cardioembolic or hypercoagulable etiology merit hospital admission.14 In many cases these etiologies may not be known at time of presentation. Evaluation, such as carotid ultrasound, may not be readily available in the ED to inform the admission decision. Several new scoring systems that utilize routine clinical features available within an hour of presentation have been developed to more objectively assess the risk of secondary stroke following a TIA. The use of these prognostic scoring systems is recommended by the National Stroke Association to aid in triaging this cohort of patients.14

Prognostic Scoring Systems

New Models of Care: An Opportunity for Hospitalists

The key to improving TIA outcomes appears to be more contingent on the speed of evaluation and initiation of appropriate therapy than on the location of the care. The EXPRESS trial studied the effect of an immediate access neurovascular clinic providing urgent evaluation and immediate treatment of nonhospitalized TIA patients versus usual care. Statistically significant reductions were seen in time to evaluation, first treatment prescription, and in 90‐day risk of recurrent stroke (10.3% versus 2.1%, P < 0.0001) after the clinic was changed to the rapid evaluation and treatment model.33

The SOS‐TIA study used a 24‐hour access hospital‐based TIA clinic to evaluate the effects of rapid assessment and interventions on hospital length of stay and clinical outcomes.34 The 90‐day stroke rate was 1.24% (95% CI, 0.712.12), which represents a 79% reduction compared to the predicted stroke rate from the ABCD² scores. With expedited evaluation and treatment, 74% of patients were able to be sent home on the same day.

The results of these 2 new studies provide compelling evidence that rapid evaluation and treatment in the first 48 hours after TIA has the potential to alter outcomes. Unfortunately not all communities have access to same day TIA clinics. Still, these findings should embolden hospitalists to advocate for urgent evaluation, such as neurology and cardiac imaging and carotid evaluation, with immediate initiation of secondary preventive therapy and early surgical intervention when appropriate. In most cases these changes will require process transformations that present prime opportunities for hospitalists to reengineer systems of care.

Incorporating Prognostic Scores into Clinical Practice

Applying the evidence to practice requires calculation of the early risk but also awareness of the community resources available. High‐risk patients with an ABCD² score of 6 or 7 have a very high 8.1% risk of stroke within the next 48 hours. Given the catastrophic outcomes frequently seen after second strokes, these patients warrant inpatient admission to facilitate the immediate initiation of appropriate secondary prevention and potentially shorten time to thrombolysis if an early stroke occurs. Intermediate‐risk patients with ABCD² scores of 4 and 5 have a 4.1% 2‐day risk of stroke and may be considered for admission, hospital observation, or expedited clinic evaluation contingent on local availability. As many as one‐third of TIA patients will be categorized as low risk with a score of 0 to 3. These patients have a 2‐day risk of stroke of only 1% and are likely safe for prompt outpatient evaluation and management. The new, validated, ABCD² score is not a substitute for individualized judgment, but is helpful in developing admission guidelines in cooperation between neurologists, emergency room physicians, and hospitalists, and in using evidence‐based medicine to provide optimal care for the patient presenting with a TIA.

Stroke and TIA arise from identical etiologies, respond to the same secondary preventive measures, and should be considered part of the spectrum of an ischemic cerebral syndrome. Recognizing TIA as a medical emergency with high rates of secondary stroke and subsequent disability allows institution of therapies with appropriate urgency. Hospitalization offers the ability to rapidly coordinate the testing and secondary prevention measures but also, for high‐risk patients, offers the opportunity to reduce the time to thrombolysis for early recurrent strokes. New, validated scoring systems such as the ABCD² score help the hospitalist to decide which patients are appropriate for admission and which can be managed in progressive and traditional outpatient settings.