Impact of physical therapy weekend coverage on length of stay in an acute care community hospital.

Impact of Physical Therapy Weekend Coverage on Length of Stay in an Acute Care Community Hospital The introduction of the prospective payment system (PPS) in 1983 by the federal government changed the Medicare reimbursement system for treatment in an acute care setting. Previously, a retrospective payment system based on the actual cost of treatment had been used. After October 1983, however, a patient's diagnosis-related group (DRG) became the guideline for reimbursement. The PPS has changed incentives for hospital management and has resulted in increased attention to reducing length of stay (LOS) as a method of controlling costs and improving hospital financial performance. Under the DRG system, a hospital receives a predetermined dollar amount per patient based on the patient's diagnosis. If the hospital can care for the patient at a cost lower than the amount of the DRG reimbursement, a financial benefit results for the hospital. Reducing the LOS while maintaining the quality of patient care is, therefore, a high priority for hospital management. The role of physical therapy in this new health care environment will be determined in part by the ability of physical therapists to help the hospital operate efficiently within the constraints of the reimbursement system. If operation of the physical therapy department contributes to a shorter LOS or reduces overall costs in other ways, the hospital's financial goals will be positively affected.

Review of the Literature

The results of the limited amount of research on the effects of the Medicare PPS on physical therapy in hospitals are contradictory. One study suggested that increased use of physical therapy results from the Medicare PPS. Other publications have suggested that the pressure for cost control can lead to reduced use of physical therapy and possibly reduced quality of care.

Hospital management increasingly must decide whether to provide physical therapy coverage on a five-day-a-week or a seven-day-a-week basis. Some evidence exists that hospitals are moving toward seven-day-a-week coverage. If seven-day-a-week coverage in physical therapy results in shorter hospital stays and if the hospital is reimbursed for these stays on a prospective payment basis, weekend physical therapy service could produce a financial benefit to the hospital. A financial benefit would result if the cost savings from the reduced LOS were greater than the additional costs of operating the physical therapy department on the weekend. If weekend physical therapy does not reduce LOS, however, the additional operating costs would have no offsetting financial benefits and might reduce hospital financial performance.

Little specific evidence exists on whether seven-day-a-week physical therapy coverage results in a shorter LOS than five-day-a-week service. One study of patients with orthopedic disorders in a major teaching hospital compared LOS and number of treatments for two groups of patients, one group treated in a unit with seven-day-a-week physical therapy coverage and a second group treated with five-day-a-week coverage in another unit of the same hospital. The results showed few differences between groups in the number of treatments or LOS. The researchers did not control for patient age or sex.

This study involved patients with orthopedic disorders or stroke in an acute care community hospital in which the physical therapy department changed from five-day-a-week to seven-day-a-week coverage. In addition to comparing patients' mean LOS before and after the change in coverage, we controlled for patient age and sex and examined the effects of weekend coverage on LOS components, specifically the time interval between hospital admission and start of physical therapy and the time interval between the patient's first and last physical therapy visits. Our main hypothesis was that the change to seven-day-a-week coverage would reduce LOS.

Method

Study Site

The study was conducted at Holyoke Hospital in Holyoke, Mass, a community of over 43,000 inhabitants. Holyoke Hospital is an acute care community hospital with a capacity of 243 licensed beds.

The physical therapy department employs six physical therapists (including a manager) and two physical therapist assistants. The caseload in the department consists mainly of outpatients. Inpatients comprise only about 20% of the total patient population.

Before October 1, 1986, the hospital provided physical therapy services on a five-day-a-week basis. Beginning on October 1, 1986, services were expanded to seven-day-a-week coverage for all impatients with certain diagnoses for a six-month duration. All impatients with one of the following diagnoses were included in this study: 1) stroke, 2) osteoarthrosis or other arthropathy, 3) fracture of the lower limb, or 4) total hip or total knee replacement.

No additional physical therapists were added to the current staff to cover Saturdays and Sundays. Each weekend day, two employees were available to treat patients from 8 AM to 12 Noon. These employees earned a free day during the week.

Subjects

The initial study sample included all inpatients with one of the above-mentioned diagnoses treated by a physical therapist and discharged from the hospital from April 1, 1986, to September 30, 1986 (Group 1) or from October 1, 1986, to March 31, 1987 (Group 2). Data were obtained retrospectively from the hospital's computerized records and from the physical therapy department's files. Patients in the four diagnostic categories represented the great majority of inpatients treated by the department. Patients with osteoarthrosis, lower limb fractures, or hip or knee replacement were grouped together and referred to as patients with orthopedic disorders.

For analytical purposes, patients were omitted from the study sample whose particular circumstances were unusual and might give a misleading picture of the experience of a "typical" patient. The following patients were omitted: 1) patients whose hospital stay included one or more days designated as administratively necessary, 2) patients whose first and last physical therapy visits did not take place in one of the designated time periods (ie, patients whose first visit was before April 1, 1986, and whose last visit was after September 30, 1986, and patients whose first visit was before October 1, 1986, and whose last visit was after March 31, 1987), and 3) patients who had several periods of physical therapy during their stay (ie, patients who were treated a few times and then were treated again, perhaps after further surgery). The totgal number of patients excluded for these reasons was about 17% of the initial study sample. The data analysis was based on a sample of 137 patients in Group 1 and 136 patients in Group 2.

Variables

We obtained data from patient records on the following variables: 1) "total LOS," or the number of days from hospital admission to hospital discharge; 2) "time from admission to initial physical therapy visit," or the number of days from hospital admission until the patient first visited the physical therapy department; 3) "time from first to last physical therapy vist," or the number of calendar days during which the patient was under the care of the physical therapy department; and 4) "number of physical therapy visits," or the number of different sessions the patient had with a therapist. Information on patients' age, sex, diagnosisM and surgical status was also recorded.

Data Analysis

Two types of analyses were conducted. We first compared the two patient groups with respect to 1) total LOS, 2) time from admission to initial physical therapy visit, 3) time from first to last physical therapy vist, and 4) number of physical therapy visits. We divided the two groups into subgroups according to diagnosis and surgical status. Although this method created very small subgroups, we believed that it would create more homogeneous groups of patients and thus more meaningful comparisons. The statistical significance of the difference between the two main groups was evaluated in two ways. At statistic was computed to test the null hypothesis that no difference existed in the mean values of LOS. In addition, because most of the distributions were non-normal, we performed a nonparametric Wilcoxon test for differences between the two groups. A 5% significance level was used for hypothesis testing.

We also performed a multivariate analysis of variance (MANOVA) of the determinants of LOS using multiple regression for two reasons. First, the multiple regression allowed us to control statistically for patient characteristics that may have affected LOS (eg, age, sex). Second, in a study such as this where the two groups represented different time periods, LOS differences may be related to various factors that change over time. For example, a national trend exists toward shorter hospital stays and substitution of outpatient for inpatient care. Hospitals are also focusing more on discharge planning and other management techniques to shorten patients' LOS. The inclusion of a trend variable in the regression equation was one way to control for such effects and to more accurately discern the effect of the main independent variable, weekend service.

Multiple regressions were performed for three dependent variables: 1) total LOS, 2) time from admission to initial physical therapy visit, and 3) time from first to last physical therapy visit. The independent variables were 1) age, 2) sex (1 if patient was male, 0 if female), 3) surgery (1 if patient had an operation, 0 otherwise), 4) number of physical therapy visits, 5) a trend variable (1 if patient was discharged in April 1986, with months numbered consecutively up to 12 for patients discharged in March 1987), and 6) a weekend-service variable (1 for patients in Group 2, 0 for Group 1). The weekend-service variable was the variable of major interest. A negative estimated regression coefficient for this variable would provide evidence that, when other factors were controlled for, LOS was shorter under weekend physical therapy service.

Results

The basic characteristics of the two groups of patients are listed in Table 1. The differences between the two groups were not pronounced, but patients in Group 2 were somewhat more likely to be men, to have an orthopedic disorder, and to have undergone surgery. The last four variables (number of physical therapy visits, total LOS, and the two segments of the total LOS) represented dependent variables in the study. The means of these variables were lower for Group 2 than for Group 1, although the medians were not. There was great variability in these data as indicated by the wide ranges and standard deviations, which were often larger than the medians and means, respectively.

Table 2 presents the results of the two-group comparison of mean total LOS. For every subgroup except patients with a nonsurgical orthopedic diagnosis, the mean LOS was shorter for patients in Group 2 than in Group 1. In only one subgroup, patients with stroke who did not have surgery, was the difference in LOS large enough to be statistically significant at the 5% level. In this subgroup, LOS was about seven days (40%) shorter for patients in Group 2 than in Group 1.

The results for LOS segments and number of physical therapy visits between groups are shown in Table 3. Patients with stroke who did not have surgery in Group 2 showed significantly less time from admission to initial physical therapy visit and from first to last physical therapy visit than the same subgroup in Group 1. This subgroup also had a lower mean number of visits in Group 2 than in Group 1. Mean time from first to last physical therapy visit was slightly higher for the other three subgroups in Group 2 than in Group 1. Mean time from admission to initial physical therapy visit was lower in Group 2 than in Group 1 for two of these three subgroups.

Table 4 summarizes the results of the multiple regression analysis. For each dependent variable, a regression analysis was performed for age, sex, surgical status, trend variable, and weekend coverage variable (equations 1, 3, 5, 6, 7, and 9). In all but one of these equations, the weekend variable was negative. The results were statistically significant in the equations for total LOS and first-to-last physical therapy visit for patients with stroke.

The other four equations controlled for the effect of number of visits. The LOS may have been shorter for Group 2 simply because those patients had fewer physical therapy visits. Equations 2, 4, 8 and 10 add the number of visits as an independent variable to the regression analysis. The fact that the weekend coverage variable coefficient remained negative and (in three of the equations) statistically significant suggests that when the number of visits was held constant, LOS was shorter under weekend service.

The other variables included in the regression analysis were not a major focus of this study. Increased age was associated with a longer LOS. Patients with stroke tended to stay longer than patients with orthopedic disorders, and patients who had surgery stayed longer than patients who did not have surgery. Sex was not significant in most equations, but its positive coefficient indicated that men tended to have a slightly longer LOS than women. The trend variable usually was positive, suggesting that when other factors were controlled for, LOS increased slightly over the 12-month period.

Discussion

The results of this study, especially the negative estimated regression coefficient for weekend coverage, supported the hypothesis that weekend coverage for physical therapy would reduce LOS. The results were very different for patients with stroke or orthopedic disorders, the two diagnostic groups examined in the study. The effect of weekend coverage for patients with stroke was statistically significant. The effect of weekend coverage for patients with orthopedic disorders was somewhat weaker statistically, and the reduction in LOS was smaller than for patients with stroke.

One implication of five-day-a-week physical therapy service is that patients who are ready to begin physical therapy on Saturday or Sunday must wait until Monday to start. With a seven-day-a-week schedule, patients who are ready to begin physical therapy on a weekend may do so. This difference in scheduling is probably the main reason for the observed reduction in time from hospital admission to first physical therapy visit.

The reasons for the effect of weekend service on the time period from first to last physical therapy visit are less clear. If a certain number of visits is needed for a patient's recovery, weekend service may enable treatment to be completed in a smaller number of calendar days. If consecutive daily visits are more effective than five-day-a-week coverage in aiding a patient's recovery, weekend service may reduce the number of visits needed. Our findings suggest that both of these effects could be present. Patients with stroke required fewer visits under weekend coverage. Even when the number of visits was controlled for, however, weekend coverage still affected the number of calendar days from first to last physical therapy visit.

The differences among the subgroups in mean values of variables as well as the effect of weekend coverage were large. Patient age and sex also had an effect, and controlling for these variables is important. Hospital management may want to consider such differences when deciding on a weekend-coverage policy. From a financial perspective, weekend coverage may be justified for some groups of patients but not for others.

The results of this study showed a greater effect of weekend service on LOS than the study of Holden and Daniele. [5] This difference may be because our diagnostic categories were different or because we controlled for certain factors not previously controlled for. A possible limitation of our study is that the comparison was between time periods and, thus, did not involve a control group of patients. A variety of time-dependent factors were only partially revealed by our statistical methodology. We also could not control for factors such as severity of illness, type of insurance coverage, bed availability in the institutions to which patients were discharged, and differences in medical and physical theraphy practice patterns that may have influenced LOS. Our results suggest that future studies should control for patient characteristics and that a multivariate analysis is necessary. In addition, it may be useful to include as dependent variables not just the total LOS but also the time segments both before and after the start of physical therapy.

Conclusion

The results of this study suggest that providing physical therapy on a seven-day-a-week basis rather than five days a week may result in a decreased LOS for patients in a community hospital. Considerable differences in this effect are associated with patient age, sex, and diagnosis.

Acknowledgments

We thank Debbie Olene for research assistance and Irene Powell, Jane McCusker, and Stephen Judd for helpful comments on a draft of this article. We also thank the physical therapy staff at Holyoke Hospital for collecting the data.

References

[1] Feder J, Hadley J, Zuckerman S: How did Medicare's prospective payment system affect hospitals? N Engl J Med 317:867-873, 1987

[2] Dore D: Effect of the Medicare prospective payment system on the utilization of physical therapy. Phys Ther 67:964-966, 1987

[3] PPS found reducing PT; cost fighting may backfire. PT Bulletin 2(31):2, 1987

[4] Hooper PJ, Dijkers M: Weekend therapy in rehab hospitals: A survey of costs and benefits. Clinical Management in Physical Therapy 7(1):16,17,21, 1987

[5' Holden MD, Daniele CA: Comparison of seven- and five-day physical therapy coverage in patients with acute orthopedic disorders. Phys Ther 67:1240-1246, 1987

J Rapoport, PhD, is Professor of Economics, Mount Holyoke College, South Hadley, MA 01075-1461 (USA).

M Judd-Van Eerd, MCed, was Manager, Physical Therapy Department, Holyoke Hospital, 575 Beech St, Holyoke, MA 01040, when this study was completed. Her current address is 566 S Catalina Blvd, Pasadena, CA 91106.

This article was submitted April 8, 1988; was with the authors for revision for two weeks; and was accepted August 15, 1988.