Lack of Timely PCP Follow‐Up

Study Setting and Population

This prospective cohort enrolled a convenience sample of patients admitted to Internal Medicine ward teams at the University of Colorado Hospital Anschutz Inpatient Pavilion between December 2007 and March 2008. Up to 2 patients were enrolled on weekdays on the morning following admission (ie, Sunday night through Thursday night admissions). Patients were screened for study entry if they were able to participate in an interview as identified by their medical team and available in their room. Of a total of 121 patients screened for study entry by a professional research assistant (PRA), 75 ultimately provided HIPAA authorization, informed consent, and completed the in‐hospital interview. The most common reasons for screened patients refusing study enrollment included being not interested (26) and too ill (10). Ten subjects were lost to follow‐up after hospital discharge, including one subject who was deceased. Therefore, 65 patients successfully completed the follow‐up phone interview and were included in the analyses. Characteristics of the 121 screened patients and the 75 study patients were similar with respect to sex, age, race, and payer mix, and representative of the demographics of the patient population at large. Case mix indices (mean) were similar among the 121 screened (1.23), 75 enrolled (1.27), and final 65 study patients (1.25).

Exclusion Criteria

Patients admitted to the medical observation unit; patients admitted at night who are ultimately reassigned to specialty services (Oncology, Cardiology, Hepatology and Acute Care for the Elderly) were excluded. Human immunodeficiency virus (HIV) patients were excluded because of routine outpatient ID follow‐up; patients <18 years of age; patients lacking a telephone; patients admitted on Friday and Saturday nights; and outside hospital transfers.

Measures

The primary study outcome was the rate of timely PCP follow‐up defined as that occurring within 4 weeks of hospital discharge. PCP was defined in this study as either a patient's known PCP (or another provider in the same clinic), or a nurse practitioner/physician assistant. Patients seen in follow‐up by a specialist related to the discharge diagnosis, eg, an Endocrinologist in a patient hospitalized for Diabetic complications; a Rheumatologist following up an SLE patient, etc., were also counted as having PCP follow‐up as defined in this study.

Additional outcomes included three measures of hospital readmission: hospital readmission for same condition; hospital readmission or other care sought (ie, ED, Urgent Care) for same condition; and hospital readmission for any condition, and index hospital LOS. The distinction between same condition and any condition was made in an attempt to delineate a potentially preventable readmission (as an example, one study patient was subsequently readmitted with a gunshot injury that clearly would not have been affected by the presence of any PCP follow‐up). Determination of same vs. any condition was made by the investigators through information obtained from patients on follow‐up phone interviews: Have you been readmitted to the University Hospital or another hospital since your discharge last month from the University Hospital? If yes: where, when, and why? The investigators determined same vs. any through comparing this information to the primary diagnosis from the index hospitalization obtained from the final discharge documentation. A condition was considered same if the readmission was for the same condition or for treatment/complications related to the index hospitalized condition.

Descriptive data collected included patient demographics, diagnoses, insurance status, presence of an identified, established PCP, time to PCP follow‐up in weeks, effects of payer source, admitting service (hospitalist vs. General Internal Medicine (GIM) attending), and nature of presenting illness (acute vs. acute on chronic condition). Categories of insurance obtained from chart review included commercial, self‐pay (uninsured), Medicare, Medicaid and Veterans.

Data Collection

A PRA screened and obtained informed consent and a Health Insurance Portability and Accountability Act (HIPAA) waiver from patients the day following admission. At that time, the PRA obtained the patients' vital information from chart review and a scripted patient interview: age, sex, PCP, categories of insurance, contact phone numbers, and admitting date and diagnoses. The in‐house interview included eight questions examining a patient's experiences of and attitudes toward PCPs. Four weeks after discharge, patients were contacted by the PRA via telephone. Scripted telephone interviews were used to determine occurrence and timing of PCP follow‐up and hospital readmission status (to any hospital) per patient self‐report. Potential barriers to PCP follow‐up were assessed. Up to 3 attempts were made to contact study subjects out to 4 weeks from the initial call (8 weeks total). If an appointment for an enrolled patient had been made, but had not yet occurred, an additional phone call was made 2 weeks later to determine whether, and when, the appointment was kept. Review of discharge summaries determined a patient's hospital LOS.

Data Analysis

Descriptive statistics were calculated for the study population. Univariate comparisons were completed for patient characteristics and study outcomes for patients with and without PCP follow‐up. We used t‐tests for continuous variables (age and LOS) and chi‐square or Fisher's exact tests when necessary for dichotomous variables (gender, uninsured vs. insured, and all hospital readmission outcomes). Comparisons according to PCP follow‐up for the categorical variables were tested with the Cochran‐Mantel‐Haenszel statistic for general association (race and insurance category) or for trends in the ordinal variable (education).

Patient characteristics and study outcomes with univariate P value < 0.1 were assessed for inclusion in the multivariate logistic regression models. Separate logistic regression models were examined with PCP follow‐up (yes/no) as the explanatory variable and the 3 hospital readmission rates as the outcomes. Final logistic regression models included the primary predictor, PCP follow‐up, along with potential predictor variables with P value < 0.05. Statistical analyses were carried out using SAS version 9.2 (SAS Institute, Cary, NC).

This protocol was approved by the Colorado Multiple Institutional Review Board (COMIRB) prior to the implemented study.

Readmissions

The 30‐day readmission rates for all study subjects were 12.3% for a patient's same medical condition, 17.2% for readmission or other care sought for the same condition, and 21.5% for any condition. Table 2 contains univariate comparisons for the patient outcomes of readmission and LOS stratified by timely PCP follow‐up. Hospital readmission for the same medical condition was significantly higher in patients lacking timely PCP follow‐up compared to those with timely PCP follow‐up (21.2% vs. 3.1%, P = 0.05). The composite outcome of hospital readmission and/or other care sought (emergency department or urgent care) for a patient's same condition was also significantly higher in patients lacking timely PCP follow‐up (28.1% vs. 6.3%; P = 0.02). However, hospital readmission for any condition did not differ with absence of timely PCP follow‐up.

Multiple logistic regression revealed that patients lacking timely PCP follow‐up were 10 times more likely to be readmitted for the same condition within 30 days of hospital discharge (odds ratio [OR] = 9.9; P = 0.04) and nearly seven times as likely to be readmitted for the same condition or receive other care (OR = 6.8, P = 0.02) (Table 3).

LOS

Overall hospital LOS in all patients was 5.4 4.6 days. In patients lacking timely PCP follow‐up, there was a trend toward longer hospital LOS: 6.3 days vs. 4.4 days, P = 0.11. For all uninsured study patients (17), the mean LOS was 6.4 days vs. 5.0 days for all other insurance categories, P = 0.31.

Insurance Status

Being uninsured was associated with a patient lacking timely PCP follow‐up (P = 0.06), but was not directly associated with higher readmission or longer hospital LOS (OR = 1.0, P = 0.96). The lack of insurance was not a significant predictor of hospital readmission in the multiple logistic regression models.

Timing of PCP Follow‐Up

In evaluating timing of any PCP follow‐up after hospital discharge and clinical outcomes, most PCP follow‐up (90.6%) occurred within the first 2 weeks following hospital discharge. However, we found no statistical difference between timing of post‐discharge PCP follow‐up and hospital readmission outcomes (hospital readmission for same reason, P = 0.51; hospital readmission or other care sought for same reason, P = 0.89), or in hospital LOS (P = 0.87). Timing of PCP follow‐upwhen comparing post‐hospitalization follow‐up <1 week, 1 to 2 weeks, and 2 to 4 weekswas not predictive of readmission rates or LOS.

Established PCP

When significance of having an established PCP prior to hospital admission was evaluated, 52 patients reported having an established PCP on hospital admission (80%), half of whom were Medicare patients. Of the 13 patients with no PCP on admission, the majority (10) were self‐pay (77%, P < 0.0001). Interestingly, only 29 (55.8%) of the patients who reported a PCP on admission to the hospital saw their PCP within 4 weeks of hospital discharge. Of 13 patients without a PCP on admission, only 3 obtained 4‐week PCP follow‐up. When we examined our study outcomes for subjects stratified by the presence of an established PCP prior to hospitalization, we found univariate association with timely post‐discharge PCP follow‐up (56% of those with established PCP vs. 23% of those without, P = 0.04), but no difference in readmission rates or hospital LOS.

Severity of patient illnessmeasured using hospital data and the case mix index (CMI)of the 3 patient populations (screened, enrolled, final) was quite similar. The CMI (mean) for the 121 screened patients was 1.23. The CMI for the 75 enrolled patients was 1.27. And the CMI in the 65 final study patients was 1.25. When evaluating illness severity (CMI) of patients in relation to hospital LOS between the 2 final study populations, the CMI (median) was also similar: 1.15 for the 32 patients with timely PCP follow‐up vs. 1.11 for the 33 patients without timely PCP follow‐up.

We found no association when looking at the rate of timely PCP follow‐up based on admitting service attending, or acute vs. acute on chronic diagnosis.

Barriers to PCP follow‐up most frequently cited by study patients were: lacking a PCP (no established PCP prior to hospital, no insurance, out of town, recently changed insurance), could not get an appointment, discharged to a half‐way house, and saw another doctor (specialist unrelated to discharge diagnosis).

Limitations

This study was performed at a single, academic institution limiting its' generalizability. In addition, this small cohort study, which took place over four winter months, may have implicit biases toward certain disease entities and follow‐up issues unique to study size and season. The small study size was dictated by a finite amount of available resources, potentially contributing to minor inconsistencies with some of the results. While statistical significance was still seen with many of our results, a much larger study may better enhance the study outcomes.

It also remains unclear why the effects of PCP follow‐up were evident for a patient's same condition, but not for any condition. The distinction between designations is potentially subjective and may be difficult to accurately determine. Most existing readmission studies in the literature assign readmission for any condition. A future, larger study may be able to examine whether this difference exists between same vs. any condition.

As an academic medical center, access to specialty clinics may be facilitated, thus increasing PCP follow‐up in patients who might otherwise not have it available to them. Additionally, our subjects were limited to a convenience sample of the population of the general medicine wards and may not be representative of all medical inpatients. Patients lacking a telephone were missed. We relied on patient recollection and self‐report of PCP follow‐up visits and re‐hospitalizations. While we acknowledge limitations of patient self‐report, both in communication and comprehension, we believe patients are reasonably able to report on whether or not they were readmitted to the hospital, the cause of their readmission and whether/when they had PCP follow‐up. Patient self‐report could be collected systematically and without long time lags. Finally, the research team did not have reliable access to readmission data for hospitals other than the facility in which the study was conducted.

It is possible patients readmitted early after discharge may have been counted as lacking PCP follow‐up simply because the readmission occurred so soon after discharge precluding the opportunity for PCP follow‐up to occur. The effects of patients having non‐PCP (home health nurse, pharmacist, phone advice) follow‐up after hospital discharge were not examined.

Also, LOS and readmission to a hospital may be more a reflection of disease severity than the absence of PCP follow‐up, ie, patients ultimately readmitted after hospital discharge may have been a sicker subset of patients upon index hospitalization.

In this urban academic medical center, discharged medicine patients commonly lack timely PCP follow‐up. The lack of timely PCP follow‐up after hospital discharge was associated with higher rates of readmission and a non‐significant trend toward longer hospital lengths of stay. Hospital discharge represents a period of significant risk in patient care necessitating the effective continuation of treatment plans including follow‐up of laboratory, radiology or other testing, and management by a variety of providers. PCPs may play a crucial role in care coordination during this period. Structured intervention performed at the time of discharge might increase post‐hospital PCP access and facilitate timely PCP follow‐up to ensure continuity of needed care after hospital discharge in the most vulnerable patients. Such interventions might include systems improvements, such as increasing PCP access in the post‐hospital period, to increase the likelihood that complex needs are met at a vulnerable period in patient care.

A more effective handoff between inpatient and outpatient settings may ultimately improve clinical outcomes, diminish resource utilization, and decrease overall healthcare costs.

Study Setting and Population

This prospective cohort enrolled a convenience sample of patients admitted to Internal Medicine ward teams at the University of Colorado Hospital Anschutz Inpatient Pavilion between December 2007 and March 2008. Up to 2 patients were enrolled on weekdays on the morning following admission (ie, Sunday night through Thursday night admissions). Patients were screened for study entry if they were able to participate in an interview as identified by their medical team and available in their room. Of a total of 121 patients screened for study entry by a professional research assistant (PRA), 75 ultimately provided HIPAA authorization, informed consent, and completed the in‐hospital interview. The most common reasons for screened patients refusing study enrollment included being not interested (26) and too ill (10). Ten subjects were lost to follow‐up after hospital discharge, including one subject who was deceased. Therefore, 65 patients successfully completed the follow‐up phone interview and were included in the analyses. Characteristics of the 121 screened patients and the 75 study patients were similar with respect to sex, age, race, and payer mix, and representative of the demographics of the patient population at large. Case mix indices (mean) were similar among the 121 screened (1.23), 75 enrolled (1.27), and final 65 study patients (1.25).

Exclusion Criteria

Patients admitted to the medical observation unit; patients admitted at night who are ultimately reassigned to specialty services (Oncology, Cardiology, Hepatology and Acute Care for the Elderly) were excluded. Human immunodeficiency virus (HIV) patients were excluded because of routine outpatient ID follow‐up; patients <18 years of age; patients lacking a telephone; patients admitted on Friday and Saturday nights; and outside hospital transfers.

Measures

The primary study outcome was the rate of timely PCP follow‐up defined as that occurring within 4 weeks of hospital discharge. PCP was defined in this study as either a patient's known PCP (or another provider in the same clinic), or a nurse practitioner/physician assistant. Patients seen in follow‐up by a specialist related to the discharge diagnosis, eg, an Endocrinologist in a patient hospitalized for Diabetic complications; a Rheumatologist following up an SLE patient, etc., were also counted as having PCP follow‐up as defined in this study.

Additional outcomes included three measures of hospital readmission: hospital readmission for same condition; hospital readmission or other care sought (ie, ED, Urgent Care) for same condition; and hospital readmission for any condition, and index hospital LOS. The distinction between same condition and any condition was made in an attempt to delineate a potentially preventable readmission (as an example, one study patient was subsequently readmitted with a gunshot injury that clearly would not have been affected by the presence of any PCP follow‐up). Determination of same vs. any condition was made by the investigators through information obtained from patients on follow‐up phone interviews: Have you been readmitted to the University Hospital or another hospital since your discharge last month from the University Hospital? If yes: where, when, and why? The investigators determined same vs. any through comparing this information to the primary diagnosis from the index hospitalization obtained from the final discharge documentation. A condition was considered same if the readmission was for the same condition or for treatment/complications related to the index hospitalized condition.

Descriptive data collected included patient demographics, diagnoses, insurance status, presence of an identified, established PCP, time to PCP follow‐up in weeks, effects of payer source, admitting service (hospitalist vs. General Internal Medicine (GIM) attending), and nature of presenting illness (acute vs. acute on chronic condition). Categories of insurance obtained from chart review included commercial, self‐pay (uninsured), Medicare, Medicaid and Veterans.

Data Collection

A PRA screened and obtained informed consent and a Health Insurance Portability and Accountability Act (HIPAA) waiver from patients the day following admission. At that time, the PRA obtained the patients' vital information from chart review and a scripted patient interview: age, sex, PCP, categories of insurance, contact phone numbers, and admitting date and diagnoses. The in‐house interview included eight questions examining a patient's experiences of and attitudes toward PCPs. Four weeks after discharge, patients were contacted by the PRA via telephone. Scripted telephone interviews were used to determine occurrence and timing of PCP follow‐up and hospital readmission status (to any hospital) per patient self‐report. Potential barriers to PCP follow‐up were assessed. Up to 3 attempts were made to contact study subjects out to 4 weeks from the initial call (8 weeks total). If an appointment for an enrolled patient had been made, but had not yet occurred, an additional phone call was made 2 weeks later to determine whether, and when, the appointment was kept. Review of discharge summaries determined a patient's hospital LOS.

Data Analysis

Descriptive statistics were calculated for the study population. Univariate comparisons were completed for patient characteristics and study outcomes for patients with and without PCP follow‐up. We used t‐tests for continuous variables (age and LOS) and chi‐square or Fisher's exact tests when necessary for dichotomous variables (gender, uninsured vs. insured, and all hospital readmission outcomes). Comparisons according to PCP follow‐up for the categorical variables were tested with the Cochran‐Mantel‐Haenszel statistic for general association (race and insurance category) or for trends in the ordinal variable (education).

Patient characteristics and study outcomes with univariate P value < 0.1 were assessed for inclusion in the multivariate logistic regression models. Separate logistic regression models were examined with PCP follow‐up (yes/no) as the explanatory variable and the 3 hospital readmission rates as the outcomes. Final logistic regression models included the primary predictor, PCP follow‐up, along with potential predictor variables with P value < 0.05. Statistical analyses were carried out using SAS version 9.2 (SAS Institute, Cary, NC).

This protocol was approved by the Colorado Multiple Institutional Review Board (COMIRB) prior to the implemented study.

Readmissions

The 30‐day readmission rates for all study subjects were 12.3% for a patient's same medical condition, 17.2% for readmission or other care sought for the same condition, and 21.5% for any condition. Table 2 contains univariate comparisons for the patient outcomes of readmission and LOS stratified by timely PCP follow‐up. Hospital readmission for the same medical condition was significantly higher in patients lacking timely PCP follow‐up compared to those with timely PCP follow‐up (21.2% vs. 3.1%, P = 0.05). The composite outcome of hospital readmission and/or other care sought (emergency department or urgent care) for a patient's same condition was also significantly higher in patients lacking timely PCP follow‐up (28.1% vs. 6.3%; P = 0.02). However, hospital readmission for any condition did not differ with absence of timely PCP follow‐up.

Multiple logistic regression revealed that patients lacking timely PCP follow‐up were 10 times more likely to be readmitted for the same condition within 30 days of hospital discharge (odds ratio [OR] = 9.9; P = 0.04) and nearly seven times as likely to be readmitted for the same condition or receive other care (OR = 6.8, P = 0.02) (Table 3).

LOS

Overall hospital LOS in all patients was 5.4 4.6 days. In patients lacking timely PCP follow‐up, there was a trend toward longer hospital LOS: 6.3 days vs. 4.4 days, P = 0.11. For all uninsured study patients (17), the mean LOS was 6.4 days vs. 5.0 days for all other insurance categories, P = 0.31.

Insurance Status

Being uninsured was associated with a patient lacking timely PCP follow‐up (P = 0.06), but was not directly associated with higher readmission or longer hospital LOS (OR = 1.0, P = 0.96). The lack of insurance was not a significant predictor of hospital readmission in the multiple logistic regression models.

Timing of PCP Follow‐Up

In evaluating timing of any PCP follow‐up after hospital discharge and clinical outcomes, most PCP follow‐up (90.6%) occurred within the first 2 weeks following hospital discharge. However, we found no statistical difference between timing of post‐discharge PCP follow‐up and hospital readmission outcomes (hospital readmission for same reason, P = 0.51; hospital readmission or other care sought for same reason, P = 0.89), or in hospital LOS (P = 0.87). Timing of PCP follow‐upwhen comparing post‐hospitalization follow‐up <1 week, 1 to 2 weeks, and 2 to 4 weekswas not predictive of readmission rates or LOS.

Established PCP

When significance of having an established PCP prior to hospital admission was evaluated, 52 patients reported having an established PCP on hospital admission (80%), half of whom were Medicare patients. Of the 13 patients with no PCP on admission, the majority (10) were self‐pay (77%, P < 0.0001). Interestingly, only 29 (55.8%) of the patients who reported a PCP on admission to the hospital saw their PCP within 4 weeks of hospital discharge. Of 13 patients without a PCP on admission, only 3 obtained 4‐week PCP follow‐up. When we examined our study outcomes for subjects stratified by the presence of an established PCP prior to hospitalization, we found univariate association with timely post‐discharge PCP follow‐up (56% of those with established PCP vs. 23% of those without, P = 0.04), but no difference in readmission rates or hospital LOS.

Severity of patient illnessmeasured using hospital data and the case mix index (CMI)of the 3 patient populations (screened, enrolled, final) was quite similar. The CMI (mean) for the 121 screened patients was 1.23. The CMI for the 75 enrolled patients was 1.27. And the CMI in the 65 final study patients was 1.25. When evaluating illness severity (CMI) of patients in relation to hospital LOS between the 2 final study populations, the CMI (median) was also similar: 1.15 for the 32 patients with timely PCP follow‐up vs. 1.11 for the 33 patients without timely PCP follow‐up.

We found no association when looking at the rate of timely PCP follow‐up based on admitting service attending, or acute vs. acute on chronic diagnosis.

Barriers to PCP follow‐up most frequently cited by study patients were: lacking a PCP (no established PCP prior to hospital, no insurance, out of town, recently changed insurance), could not get an appointment, discharged to a half‐way house, and saw another doctor (specialist unrelated to discharge diagnosis).

Limitations

This study was performed at a single, academic institution limiting its' generalizability. In addition, this small cohort study, which took place over four winter months, may have implicit biases toward certain disease entities and follow‐up issues unique to study size and season. The small study size was dictated by a finite amount of available resources, potentially contributing to minor inconsistencies with some of the results. While statistical significance was still seen with many of our results, a much larger study may better enhance the study outcomes.

It also remains unclear why the effects of PCP follow‐up were evident for a patient's same condition, but not for any condition. The distinction between designations is potentially subjective and may be difficult to accurately determine. Most existing readmission studies in the literature assign readmission for any condition. A future, larger study may be able to examine whether this difference exists between same vs. any condition.

As an academic medical center, access to specialty clinics may be facilitated, thus increasing PCP follow‐up in patients who might otherwise not have it available to them. Additionally, our subjects were limited to a convenience sample of the population of the general medicine wards and may not be representative of all medical inpatients. Patients lacking a telephone were missed. We relied on patient recollection and self‐report of PCP follow‐up visits and re‐hospitalizations. While we acknowledge limitations of patient self‐report, both in communication and comprehension, we believe patients are reasonably able to report on whether or not they were readmitted to the hospital, the cause of their readmission and whether/when they had PCP follow‐up. Patient self‐report could be collected systematically and without long time lags. Finally, the research team did not have reliable access to readmission data for hospitals other than the facility in which the study was conducted.

It is possible patients readmitted early after discharge may have been counted as lacking PCP follow‐up simply because the readmission occurred so soon after discharge precluding the opportunity for PCP follow‐up to occur. The effects of patients having non‐PCP (home health nurse, pharmacist, phone advice) follow‐up after hospital discharge were not examined.

Also, LOS and readmission to a hospital may be more a reflection of disease severity than the absence of PCP follow‐up, ie, patients ultimately readmitted after hospital discharge may have been a sicker subset of patients upon index hospitalization.

In this urban academic medical center, discharged medicine patients commonly lack timely PCP follow‐up. The lack of timely PCP follow‐up after hospital discharge was associated with higher rates of readmission and a non‐significant trend toward longer hospital lengths of stay. Hospital discharge represents a period of significant risk in patient care necessitating the effective continuation of treatment plans including follow‐up of laboratory, radiology or other testing, and management by a variety of providers. PCPs may play a crucial role in care coordination during this period. Structured intervention performed at the time of discharge might increase post‐hospital PCP access and facilitate timely PCP follow‐up to ensure continuity of needed care after hospital discharge in the most vulnerable patients. Such interventions might include systems improvements, such as increasing PCP access in the post‐hospital period, to increase the likelihood that complex needs are met at a vulnerable period in patient care.

A more effective handoff between inpatient and outpatient settings may ultimately improve clinical outcomes, diminish resource utilization, and decrease overall healthcare costs.

Study Setting and Population

This prospective cohort enrolled a convenience sample of patients admitted to Internal Medicine ward teams at the University of Colorado Hospital Anschutz Inpatient Pavilion between December 2007 and March 2008. Up to 2 patients were enrolled on weekdays on the morning following admission (ie, Sunday night through Thursday night admissions). Patients were screened for study entry if they were able to participate in an interview as identified by their medical team and available in their room. Of a total of 121 patients screened for study entry by a professional research assistant (PRA), 75 ultimately provided HIPAA authorization, informed consent, and completed the in‐hospital interview. The most common reasons for screened patients refusing study enrollment included being not interested (26) and too ill (10). Ten subjects were lost to follow‐up after hospital discharge, including one subject who was deceased. Therefore, 65 patients successfully completed the follow‐up phone interview and were included in the analyses. Characteristics of the 121 screened patients and the 75 study patients were similar with respect to sex, age, race, and payer mix, and representative of the demographics of the patient population at large. Case mix indices (mean) were similar among the 121 screened (1.23), 75 enrolled (1.27), and final 65 study patients (1.25).

Exclusion Criteria

Patients admitted to the medical observation unit; patients admitted at night who are ultimately reassigned to specialty services (Oncology, Cardiology, Hepatology and Acute Care for the Elderly) were excluded. Human immunodeficiency virus (HIV) patients were excluded because of routine outpatient ID follow‐up; patients <18 years of age; patients lacking a telephone; patients admitted on Friday and Saturday nights; and outside hospital transfers.

Measures

The primary study outcome was the rate of timely PCP follow‐up defined as that occurring within 4 weeks of hospital discharge. PCP was defined in this study as either a patient's known PCP (or another provider in the same clinic), or a nurse practitioner/physician assistant. Patients seen in follow‐up by a specialist related to the discharge diagnosis, eg, an Endocrinologist in a patient hospitalized for Diabetic complications; a Rheumatologist following up an SLE patient, etc., were also counted as having PCP follow‐up as defined in this study.

Additional outcomes included three measures of hospital readmission: hospital readmission for same condition; hospital readmission or other care sought (ie, ED, Urgent Care) for same condition; and hospital readmission for any condition, and index hospital LOS. The distinction between same condition and any condition was made in an attempt to delineate a potentially preventable readmission (as an example, one study patient was subsequently readmitted with a gunshot injury that clearly would not have been affected by the presence of any PCP follow‐up). Determination of same vs. any condition was made by the investigators through information obtained from patients on follow‐up phone interviews: Have you been readmitted to the University Hospital or another hospital since your discharge last month from the University Hospital? If yes: where, when, and why? The investigators determined same vs. any through comparing this information to the primary diagnosis from the index hospitalization obtained from the final discharge documentation. A condition was considered same if the readmission was for the same condition or for treatment/complications related to the index hospitalized condition.

Descriptive data collected included patient demographics, diagnoses, insurance status, presence of an identified, established PCP, time to PCP follow‐up in weeks, effects of payer source, admitting service (hospitalist vs. General Internal Medicine (GIM) attending), and nature of presenting illness (acute vs. acute on chronic condition). Categories of insurance obtained from chart review included commercial, self‐pay (uninsured), Medicare, Medicaid and Veterans.

Data Collection

A PRA screened and obtained informed consent and a Health Insurance Portability and Accountability Act (HIPAA) waiver from patients the day following admission. At that time, the PRA obtained the patients' vital information from chart review and a scripted patient interview: age, sex, PCP, categories of insurance, contact phone numbers, and admitting date and diagnoses. The in‐house interview included eight questions examining a patient's experiences of and attitudes toward PCPs. Four weeks after discharge, patients were contacted by the PRA via telephone. Scripted telephone interviews were used to determine occurrence and timing of PCP follow‐up and hospital readmission status (to any hospital) per patient self‐report. Potential barriers to PCP follow‐up were assessed. Up to 3 attempts were made to contact study subjects out to 4 weeks from the initial call (8 weeks total). If an appointment for an enrolled patient had been made, but had not yet occurred, an additional phone call was made 2 weeks later to determine whether, and when, the appointment was kept. Review of discharge summaries determined a patient's hospital LOS.

Data Analysis

Descriptive statistics were calculated for the study population. Univariate comparisons were completed for patient characteristics and study outcomes for patients with and without PCP follow‐up. We used t‐tests for continuous variables (age and LOS) and chi‐square or Fisher's exact tests when necessary for dichotomous variables (gender, uninsured vs. insured, and all hospital readmission outcomes). Comparisons according to PCP follow‐up for the categorical variables were tested with the Cochran‐Mantel‐Haenszel statistic for general association (race and insurance category) or for trends in the ordinal variable (education).

Patient characteristics and study outcomes with univariate P value < 0.1 were assessed for inclusion in the multivariate logistic regression models. Separate logistic regression models were examined with PCP follow‐up (yes/no) as the explanatory variable and the 3 hospital readmission rates as the outcomes. Final logistic regression models included the primary predictor, PCP follow‐up, along with potential predictor variables with P value < 0.05. Statistical analyses were carried out using SAS version 9.2 (SAS Institute, Cary, NC).

This protocol was approved by the Colorado Multiple Institutional Review Board (COMIRB) prior to the implemented study.

Readmissions

The 30‐day readmission rates for all study subjects were 12.3% for a patient's same medical condition, 17.2% for readmission or other care sought for the same condition, and 21.5% for any condition. Table 2 contains univariate comparisons for the patient outcomes of readmission and LOS stratified by timely PCP follow‐up. Hospital readmission for the same medical condition was significantly higher in patients lacking timely PCP follow‐up compared to those with timely PCP follow‐up (21.2% vs. 3.1%, P = 0.05). The composite outcome of hospital readmission and/or other care sought (emergency department or urgent care) for a patient's same condition was also significantly higher in patients lacking timely PCP follow‐up (28.1% vs. 6.3%; P = 0.02). However, hospital readmission for any condition did not differ with absence of timely PCP follow‐up.

Multiple logistic regression revealed that patients lacking timely PCP follow‐up were 10 times more likely to be readmitted for the same condition within 30 days of hospital discharge (odds ratio [OR] = 9.9; P = 0.04) and nearly seven times as likely to be readmitted for the same condition or receive other care (OR = 6.8, P = 0.02) (Table 3).

LOS

Overall hospital LOS in all patients was 5.4 4.6 days. In patients lacking timely PCP follow‐up, there was a trend toward longer hospital LOS: 6.3 days vs. 4.4 days, P = 0.11. For all uninsured study patients (17), the mean LOS was 6.4 days vs. 5.0 days for all other insurance categories, P = 0.31.

Insurance Status

Being uninsured was associated with a patient lacking timely PCP follow‐up (P = 0.06), but was not directly associated with higher readmission or longer hospital LOS (OR = 1.0, P = 0.96). The lack of insurance was not a significant predictor of hospital readmission in the multiple logistic regression models.

Timing of PCP Follow‐Up

In evaluating timing of any PCP follow‐up after hospital discharge and clinical outcomes, most PCP follow‐up (90.6%) occurred within the first 2 weeks following hospital discharge. However, we found no statistical difference between timing of post‐discharge PCP follow‐up and hospital readmission outcomes (hospital readmission for same reason, P = 0.51; hospital readmission or other care sought for same reason, P = 0.89), or in hospital LOS (P = 0.87). Timing of PCP follow‐upwhen comparing post‐hospitalization follow‐up <1 week, 1 to 2 weeks, and 2 to 4 weekswas not predictive of readmission rates or LOS.

Established PCP

When significance of having an established PCP prior to hospital admission was evaluated, 52 patients reported having an established PCP on hospital admission (80%), half of whom were Medicare patients. Of the 13 patients with no PCP on admission, the majority (10) were self‐pay (77%, P < 0.0001). Interestingly, only 29 (55.8%) of the patients who reported a PCP on admission to the hospital saw their PCP within 4 weeks of hospital discharge. Of 13 patients without a PCP on admission, only 3 obtained 4‐week PCP follow‐up. When we examined our study outcomes for subjects stratified by the presence of an established PCP prior to hospitalization, we found univariate association with timely post‐discharge PCP follow‐up (56% of those with established PCP vs. 23% of those without, P = 0.04), but no difference in readmission rates or hospital LOS.

Severity of patient illnessmeasured using hospital data and the case mix index (CMI)of the 3 patient populations (screened, enrolled, final) was quite similar. The CMI (mean) for the 121 screened patients was 1.23. The CMI for the 75 enrolled patients was 1.27. And the CMI in the 65 final study patients was 1.25. When evaluating illness severity (CMI) of patients in relation to hospital LOS between the 2 final study populations, the CMI (median) was also similar: 1.15 for the 32 patients with timely PCP follow‐up vs. 1.11 for the 33 patients without timely PCP follow‐up.

We found no association when looking at the rate of timely PCP follow‐up based on admitting service attending, or acute vs. acute on chronic diagnosis.

Barriers to PCP follow‐up most frequently cited by study patients were: lacking a PCP (no established PCP prior to hospital, no insurance, out of town, recently changed insurance), could not get an appointment, discharged to a half‐way house, and saw another doctor (specialist unrelated to discharge diagnosis).

Limitations

This study was performed at a single, academic institution limiting its' generalizability. In addition, this small cohort study, which took place over four winter months, may have implicit biases toward certain disease entities and follow‐up issues unique to study size and season. The small study size was dictated by a finite amount of available resources, potentially contributing to minor inconsistencies with some of the results. While statistical significance was still seen with many of our results, a much larger study may better enhance the study outcomes.

It also remains unclear why the effects of PCP follow‐up were evident for a patient's same condition, but not for any condition. The distinction between designations is potentially subjective and may be difficult to accurately determine. Most existing readmission studies in the literature assign readmission for any condition. A future, larger study may be able to examine whether this difference exists between same vs. any condition.

As an academic medical center, access to specialty clinics may be facilitated, thus increasing PCP follow‐up in patients who might otherwise not have it available to them. Additionally, our subjects were limited to a convenience sample of the population of the general medicine wards and may not be representative of all medical inpatients. Patients lacking a telephone were missed. We relied on patient recollection and self‐report of PCP follow‐up visits and re‐hospitalizations. While we acknowledge limitations of patient self‐report, both in communication and comprehension, we believe patients are reasonably able to report on whether or not they were readmitted to the hospital, the cause of their readmission and whether/when they had PCP follow‐up. Patient self‐report could be collected systematically and without long time lags. Finally, the research team did not have reliable access to readmission data for hospitals other than the facility in which the study was conducted.

It is possible patients readmitted early after discharge may have been counted as lacking PCP follow‐up simply because the readmission occurred so soon after discharge precluding the opportunity for PCP follow‐up to occur. The effects of patients having non‐PCP (home health nurse, pharmacist, phone advice) follow‐up after hospital discharge were not examined.

Also, LOS and readmission to a hospital may be more a reflection of disease severity than the absence of PCP follow‐up, ie, patients ultimately readmitted after hospital discharge may have been a sicker subset of patients upon index hospitalization.

In this urban academic medical center, discharged medicine patients commonly lack timely PCP follow‐up. The lack of timely PCP follow‐up after hospital discharge was associated with higher rates of readmission and a non‐significant trend toward longer hospital lengths of stay. Hospital discharge represents a period of significant risk in patient care necessitating the effective continuation of treatment plans including follow‐up of laboratory, radiology or other testing, and management by a variety of providers. PCPs may play a crucial role in care coordination during this period. Structured intervention performed at the time of discharge might increase post‐hospital PCP access and facilitate timely PCP follow‐up to ensure continuity of needed care after hospital discharge in the most vulnerable patients. Such interventions might include systems improvements, such as increasing PCP access in the post‐hospital period, to increase the likelihood that complex needs are met at a vulnerable period in patient care.

A more effective handoff between inpatient and outpatient settings may ultimately improve clinical outcomes, diminish resource utilization, and decrease overall healthcare costs.