These analyses provide important information for understanding who uses PAC services, how their likelihood of using them differs by certain characteristics, and which factors are most important in predicting hospital length of stay, probability of PAC use, relative probability of PAC site of care choices, hospital readmission rates, and average episode payments. Age and severity of illness factors were important in all the models.
In this work, we tested several measures of severity, including the APR-DRGs which assign severity levels based on having certain comorbid conditions; MS-DRGs which match specific primary and secondary conditions to assign a severity level; and HCC indicators of comorbid conditions. These three methods varied in their explanatory power although the models that explained the most variation in payments (Table 3-27) included both the MS-DRG and the HCC flags (28.0 percent explained variance compared to 13.0 percent for the APR-DRG and 6.8 percent for the MS-DRG-based models). Individual HCC flags varied in their impact on length of stay. Certain comorbid conditions, such as cardio-respiratory failure and shock and major complication of medical care and trauma had much larger effects than other conditions such as vascular disease (Table 3-26).
Severity of illness was important for predicting expected length of stay in the acute hospital prior to PAC use. Greater severity was associated with longer length stay, as expected, regardless of measure used. The APR-DRG model (Table 3-24) had the greatest explanatory power in these models, explaining almost 22.0 percent of the variation in hospital LOS. The MS-DRG models (Table 3-24) explained 11.0 percent of the variation but this increased to 17.0 percent when the HCC flags were added. It was also important for explaining the probability of PAC use and the type of PAC service used. Patients with higher severity scores were more likely to use LTCHs, followed by SNFs, then IRFs, and last home health services relative to outpatient therapy services.
Organizational relationships were also important for predicting use. Having a subprovider was associated with having longer acute lengths of stay suggesting that PAC units are more commonly associated with hospitals treating sicker populations. We also found greater likelihoods of using a type of PAC if the hospital had a subprovider or co-located PAC provider of that type. For example, the model showed a greater likelihood of using IRFs if the hospital has a subprovider or co-located IRF. Having a particular type of subprovider appeared to be related to the use of that provider type, with the exception of home health. SNF effects were lower than IRF or LTCH and this may be due to the availability of these providers.
Severity and organizational relationships were important for predicting readmission rates and average episode payments. The probability of readmission increased as severity increased and having a subprovider was negatively associated with readmission rates. For example, the results of the binomial regressions predicting readmission using the MS-DRG severity measures and HCCs (Set 3, Table 3-27) indicated that the odds ratio on the MS-DRG severity CC variable was 1.293 and the odds ratio on the MS-DRG severity MCC variable was 1.387 demonstrating increasing odds of readmission with increasing severity compared to beneficiaries with MS-DRG severity w/o CC. In looking at the subprovider variable in the same model, beneficiaries discharged from an acute hospital with any subprovider had 0.973 times the odds of readmission compared to beneficiaries discharged from an acute hospital without any subprovider. The odds ratio on the variable indicating the presence of any colocated provider was not statistically significant. Both factors severity and organizational relationships were also significantly associated with episode payments; as severity increased, so did the average payment per episode. Similarly, average episode payments were higher for beneficiaries treated in hospitals with PAC providers. This may reflect different resource mixes of the hospitals or reflect higher likelihood of using subproviders where they exist, all else equal.
These analyses provide a useful insight into the role of comorbid conditions in PAC utilization, the impact of patient severity as measured by both the APR-DRG and the MS-DRG on patterns of PAC use, and a more complete view of a beneficiary's total cost of care within in an episode, given the addition of the Part B, DME, and hospice data to this year's analyses. Another important contribution of this effort is the examination of the role of organizational relationships beyond first site of PAC care. Although previous work focused exclusively on the relationship between acute settings and first site of PAC, this work also looked at the effect of the presence of organizational relationships between providers across multiple settings of PAC settings.
While this information is useful for predicting episode use and payment variation, this work has also been helpful in considering the patterns of care in the Medicare program and how the mix of services may vary depending on the patients' complexity and the resources available in their local market area. The pattern analysis discussed in Section 3.7 helps us understand the way services are combined to treat individual patients. Of the 35.2 percent of hospital discharges to PAC, 52.0 percent of them go on to use additional services after the first PAC site. The episode payments and length of stay vary extensively depending on the extent to which higher cost institutional services are part of the episode or longer lasting, ambulatory services, such as home health or outpatient therapy. As shown in Figure 3-3, among the most common PAC discharge group (SNFs which admitted 41.0 percent of PAC users), average payment per episode was $8,759 but ranged from additional amounts of $3,544 for discharge to home health to an additional $29,118 for cases discharged from the SNF to LTCHs. Further, seeing how these patterns varied for medical versus rehabilitation cases was also useful for considering expected care trajectories and costs.
Analysis of the physician's role will also be important. The physician is a key player in the Medicare program, yet often PAC populations have numerous physicians involved. First, the patient's primary care provider typically sent them to the hospital. Once there, they may be seen by an emergency room physician, a hospitalist, or a surgeon. If they are discharged to PAC, they will be seen by yet another set of physicians, and depending on the type of setting, possibly more than one physician, including physiatrists, pulmonologists, infectious disease specialists, internists, and many others. Information about their various treatments may or may not go back to their primary care physician.
The data presented in Section 3.8 helps us begin to understand the physician's role in an episode of care. More work is needed on this to explore the number of physicians and place and timing of interventions across an episode of care, and further, how these may vary by different types of patients (those with different conditions or at different levels of severity of illness).
The results of the analyses in this report demonstrate the importance of understanding case complexity, particularly the patient's medical complexity. Second, comorbidity plays an important role in understanding expected costs and use, since both the type of comorbidity (as measured using the HCCs) and the number of comorbidities affect PAC utilization. We observed that the episode payments and length of stay increased with increasing numbers of comorbidities. In looking at the presence of selected comorbidities by themselves, and then in combination, we also saw a noticeable increase in episode payments in beneficiaries with combinations of selected HCCs. These analyses provide a baseline understanding of how comorbidities may affect episodes of PAC. These issues will be explored further in future work as we consider the use of the Chronic Condition Warehouse data and alternative definitions of post-acute episodes in the coming year.
The use of the MS-DRG severity measures in this year's work provided an opportunity to learn more about the severity definitions in the MS-DRGs and provided good information on ways to improve its usefulness in stratifying patients based on severity. The addition of the HCC flags was helpful for considering a person's constellation of conditions which may be complicating the use of any one service. The analyses also compared the effects of the two condition-specific severity measures. While the MS-DRGs have fewer severity levels, the two methodologies are not directly comparable. The MS-DRG system bases complexity on a condition-specific number of complicating levels while the APR-DRG system assigns severity based on the presence of specific comorbid conditions. Despite these differences in measures, both systems showed the same basic patterns of longer length episodes and higher payments being associated with more complicated MS-DRG and APR-DRG levels. The inclusion of the HCC indicator variables in the multivariate models using the MS-DRG severity measure allowed for a more specific understanding of the role of comorbidity and severity on post-acute care utilization.
The inclusion of the Part B, DME, and hospice claims in this year's episode files was a major advancement in the post-acute care analysis. Previous analyses have focused solely on the inpatient post-acute, home health, and hospital outpatient therapy settings of care, but as the results of the analyses indicate, Part B physician services are an important component of overall post-acute care payments. Across DRGs and severity levels, Part B physician services accounted for over 11.9 percent of total episode payments (Figure 3-8). Again, this work serves as a base to further work looking into physician utilization during post-acute care episodes. Of particular interest is the point at which these services occur during the post-acute episode. Do they occur primarily during the index acute hospitalization or are they follow-up visits? What types of physician specialists are treating beneficiaries in the different DRGs? Are the visits related to the index DRG or are they unrelated? These questions will be the basis of future work in this area.
Last year's report on post-acute care episodes focused on the presence of organizational relationships and the effects of organizational relationships on discharge to first site of PAC, readmission to PAC, index admission length of stay, and any post-acute use. In recognition of the fact that episodes of post-acute care are complex and often involve more than one site of care, this year's work attempted to learn more about the relationships between providers within post-acute care episodes with multiple providers. The analysis of episode patterns including acute index hospitalization, SNF, and HHA claims (episode pattern=ASH); and episodes including acute index hospitalization, IRF, and HHA claims (episode pattern=AIH) demonstrated that episodes occurring within sets of related providers were more likely to have lower episode lengths of stay and payments, after controlling for beneficiary demographics and severity. This finding suggests that there are differences in patterns of utilization for beneficiaries receiving care within providers with formal and informal organizational relationships.