Our findings reveal that hospitals in the highest accreditation category have a higher case mix index than those in lower categories. This is not a surprising finding since one would expect a higher accredited hospital, being better resourced, would admit more complex patients and thus have a higher CMI. The literature on the relation of accreditation with case mix index is not well established, in particular considering the differing components of accreditation systems, however increased technical capability has been found to be associated with higher hospital case mix index [16]. However the absence of a difference in CMI between the remaining three accreditation categories suggests that the payment system used by the MoPH which links reimbursement solely to accreditation is not appropriate. The presence of a few hospitals within categories (A and C) with considerably higher CMI than others in the same category implies that hospitals within the same category are not homogenous. Therefore the current system is unfair and induces inefficiency among and within accreditation categories.
One factor that could have especially contributed to the variation seen in hospital CMI is the lack of financial incentives for improving hospital performance. Instead the system has been built to provide hospitals with clear incentives for enhancing healthcare quality in terms of accreditation requirements.
We also found that hospitals with more than 100 beds have a higher CMI than hospitals with less than 50 beds, which may be another reflection of larger hospitals having greater technical capabilities and thus admitting more complex admissions. It is important to note that our findings regarding case mix index and hospital accreditation remained unchanged even after adjusting for hospital size and volume.
The CMI of private hospitals was higher than that of public hospitals, however the range of CMI was wider among private hospitals, with some admitting lower CMI than public hospitals. This suggests that some private hospitals may be shifting resource-intensive patients towards public hospitals, and noting that public hospitals are predominantly in remote areas and most are considered front-line in nature. In the literature case mix index is often variable between private and public hospitals, but tends to be higher in private hospitals, although the opposite may apply depending on the diagnosis groups involved [17, 18] Our findings also confirm that, though private hospitals have more advanced structures and technology and can admit higher patient case mix, they are not necessarily more efficient than public hospitals, and confirms the findings of a systemic review on low- and middle-income countries [19].
The correlation between principal diagnosis-based case mix index (CMI-ICDC) and surgical procedure-based case mix index (CMI-CPTC) is overall fair, but it is only among hospitals with more than 100 beds that ICD-CPTC is a good substitute for CMI-ICDC. A principal diagnosis-based CMI can be regarded as a more appropriate measure for hospital CMI, since it considers all admissions to a hospital rather than only surgical ones. A surgical procedure-based CMI favors hospitals with high proportion of surgeries at the expense of those with low proportion. That both CMIs are more closely correlated in larger hospitals is likely due to these hospitals having a larger proportion of their admissions that are surgical cases, and therefore a CMI-CPTC can substitute better for CMI-ICDC. Nevertheless for most hospitals CMI-CPTC would be a less fair assessment of case mix index than CMI-ICDC, and its usefulness is more as a separate additional measure, whereas principal diagnosis-based case mix index is the better measure for the aggregate complexity of patients admitted to a hospital.
Both CPT complexity and ICU/Ventilator use are fair proxies for CMI, suggesting the possibility to use the first two measures as intermediate measures of CMI, or more ideally as separate measures altogether, which could inform contracting as hospital process indicators. The lack of any correlation between surgical proportion and case mix essentially means that a hospital with more surgeries does not translate into a more complex hospital. This may be interesting to explore further, since surgical procedures are less prone to bill inflation and abuse, being based on a flat-rate fee per procedure rather than the open rate of medical admissions. While such cost-containment methods are favorable, it suggests the relation between flat-rate and case mix is a complex one.
In the literature hospital readmissions tend to increase with increasing case mix [20, 21]. We used two different measures of 30-day readmissions to capture the varying perspectives of this outcome measure, taking into consideration the local hospitalization practices of hospitals in Lebanon. Both measures yielded similar results, confirming the suggested interpretations. Readmission increased as accreditation category increased, with the lowest accreditation category (D) having lower readmissions than each of category A and C hospitals, though no significant difference was found between these latter two categories. This suggests the relation between readmission and case mix is also a complex one, where readmissions in higher accredited hospitals may be explained by more complex admissions, while readmissions in lower accredited hospitals may rather reflect poorer healthcare quality. In general for both readmission measures hospitals with more than 100 beds had higher readmission than smaller hospitals.
The absence of any difference in readmission between private and public hospitals, even for the any-hospital readmission, suggests that the healthcare quality of public hospitals is not necessarily lower than that of private hospitals, though they are intended as front-line hospitals and expectedly have lower technical capabilities.
It is important to note that caution should always be exercised in linking hospital performance indicators to reimbursement rate. Early evidence from 250 US hospitals showing Pay for Performance (P4P) increased process-quality measures by 3-4% was encouraging and has been used to inform health system reform in the US [22]. However with longer follow up these gains were decreased or lost, and other research has also found P4P to not improve patient outcomes in the long-term [23, 24]. More encouraging is recent research from England where P4P in one region resulted in reduction of mortality [25]. As noted by Epstein [26] in the long-term P4P can increase healthcare quality, but the rate at which this occurs depends on the measures and incentives placed.
Strengths
There is very limited published research from Lebanon and the Middle East – North Africa (MENA) region on reimbursement, case mix and linkages to contractual mechanisms. To our knowledge this is the first time a case mix index is developed for hospitals in Lebanon using an evidence-based methodology. This is also a first for the MoPH in making such an extensive use of its administrative database, and the process involved in implementing this research highlighted several aspects for potential improvement of the database for future use. In addition, the dataset used for all analysis in this research is a very recent one (June 2011-May 2012) and could be used to inform the upcoming MoPH hospital contracting process. The calculation of hospital CMI and ICD weights is relatively easy to perform, and in the case of the latter, more accurate weights for low-volume codes can be obtained as additional cases collected across two or more years.
Limitations
A system which heavily relies on hospital coding of cases is always prone to coding malpractice [27]. However in our research where current reimbursement is not directly linked to ICD code, there is no clear financial incentive for hospitals to alter coding. The presence of trained MoPH physician controllers that are locally placed at hospitals whose tasks include oversight of codes recorded, as well as proper medical archiving being a criterion for accreditation, would expectedly minimize such miscoding, though some miscoding will continue to occur regardless of control mechanisms. Another limitation is that we have used cost-data in terms of what the MoPH has to reimburse hospitals, and this does not capture the real cost of treating a patient borne by the hospital.