Explanatory Variables. I measured use of CMPs as a vector of the five key care management domains. First, each hospital received a composite score representing its overall use of CMPs. Composite scores were calculated by dividing the total number of processes used by the hospital across all domains by the total number of processes specified on the survey. A total of 29 CMPs were considered across all five domains (See Appendix Table 1). Next, I considered evidence from Xxxxx and colleague’s examination of patient- centered medical homes, which suggests different practices across different care domain yield differences in outcomes (Xxxxx, Saynisch, & Xxxxx-XxXxxxxx, 2016). Accordingly, I calculated individual domain scores based upon the degree of utilization to further differentiate hospital-level care management strategies. I adapted Berwolitz et al.’s (2003) methodology for quality improvement implementation to score degree of CMP implementation. Correspondingly, a hospital would receive a score for an individual CMP ranging from 0-4 depending on the degree of use as indicated on the Care Systems & Payment survey: 0 for “Not used at all”, 1 for “Used minimally”, 2 for “Used moderately”, 3 for “Used widely”, and 4 for “Used hospital-wide”. The scoring scale was validated for internal consistency using Cronbach’s alpha (Xxxxx & Xxxxxx, 1997). Then, I used individual CMP implementation scores to calculate factor scores (DiStefano, Zhu, & Mindrila, 2009; Xxxxxxx & Xxxxx, 1964) for each individual CMP domain. For both hospital composite and domain factor scores, higher score values indicate greater use of care management practices. Dependent Variables I evaluated hospital value-based performance using MSPB, HCAPHS, Hospital Readmission Rates, VBP, and Clinical Care Outcomes scores reported by Hospital Compare. MSPB and Hospital Readmissions are reported as ratios relative to 1. Ratios greater than 1 indicate excess readmissions or spending. More specifically, a MSPB value of 1.1 indicates that the hospital has average risk-adjusted spending levels 10 % higher than the median MSPB episode. HCAPHS, VBP, and clinical care outcome scores are reported as a numeric value ranging from 0 to 100 where greater scores indicate greater performance. I consider total performance score for the VBP program and the unweighted VBP sub- scores for the clinical care outcomes domain.
Explanatory Variables. In this work we considered a set of 14 bioclimatic variables commonly used in ecological modelling (see Table 1). The main advantage of these bioclimatic variables over other seasonal indices is that they are calculated and applied independently of the country’s hemisphere. In this study, we used only bioclimatic variables which could be calculated using mean temperature and precipitation data, as minimum and maximum temperatures, which allow for the calculation of other indices, were not available. In order to avoid redundancy, we eliminated from the analysis bioclimatic variables which yielded correlation values above
Explanatory Variables. CDFs. The presence or absence of CDFs is the principal explanatory variable of interest. It is conceived of as a dichotomous variable (CDFs), coded 1 if an incumbent deployed CDFs at any point during a conflict, and 0 otherwise. For each observation in the sample, I rely on individual case histories to identify whether or not an auxil- iary force was present that meets the criteria specified earlier: having been armed by the incumbent, its rank-and-file largely consisting of civilians native to the region in which they are deployed, and for the most part being tasked with defensive operations such as nightwatch and patrol duties. With this definition in mind, CDFs appear to have been deployed in 33 out of the 86 insurgencies that have started and concluded between 1945 to 2006. A simple bivariate analysis suggests that their deployment is rather weakly correlated with COIN outcomes (r=0.10).
