The aim of this study was to investigate whether psychiatric comorbidity predicts direct and indirect cost in back pain patients undergoing disc surgery and to estimate other predictors for direct and indirect costs, in a prospective study design. Summarized, we found that direct as well as indirect costs were strongly predicted by time and clinical variables including psychiatric comorbidity.
We found that psychiatric comorbidity is associated with increased direct costs, though we found almost no mental health care utilization at all measurements, i.e. psychiatric comorbidity increased non-mental health care utilization, which has also been found in other diseases [8
]. There are several possible explanations for this finding. One may be, that psychiatric comorbidity is often either not diagnosed or not treated with psychiatric services, as shown in other studies [36
]. Another possible explanation may be underreporting of this specific type of health care due to fear of stigma. Finally, stigma may also be a reason why patients do not utilize psychiatric treatment.
As a result of the therapeutic pathway, time was a very strong predictor of costs. At T0 direct costs were strongly influenced by acute disease and in particular by disc surgery resulting in the highest three month health care costs of all measurements. Acute disease and disc surgery were both also associated with sickness absence causing high indirect costs at T0. At T1 direct costs primarily resulted from inpatient rehabilitation and subsequent outpatient treatments, resulting in considerably lower direct costs compared to T0. Indirect costs strongly increased at T1 due to more sickness absence days resulting from postoperative sick leave and inpatient rehabilitation which lasted several weeks. At T2 acute treatment and rehabilitation had been completed and direct costs preliminary consisted of outpatient treatment costs, resulting in the lowest three month health care costs of all periods analysed. Accordingly, most patients were back to work again and sickness absence (and indirect costs) declined considerably.
Though mostly not significant, the interaction term of time and psychiatric comorbidity showed an identical interesting course over time for direct and indirect costs: at T0 patients with psychiatric comorbidity showed higher costs than patients without; at T1 this difference diminished, but was found again at T2. In our opinion, this course could be seen as an artefact resulting from the rather strictly organised therapy flow during the 3-month interval preceding the T1 assessment. This interval was characterized by the postoperative therapy and an inpatient rehabilitation lasting mostly three to four weeks, which both are highly standardized. On the one hand this limited patient’s choice of health care utilization and thus equalized direct costs of patients with and without psychiatric comorbidity. On the other hand, relative standardized durations of postoperative sick leave and rehabilitation, may also have equalized indirect costs of patients with and without psychiatric comorbidity. However, at T2 this “effect” of therapy was not present anymore; hence the impact of psychiatric comorbidity on direct as well as indirect costs was observable again. In conclusion the impact of psychiatric comorbidity on cost seems to have persisted after disc surgery and rehabilitation.
Besides time and psychiatric comorbidity, direct and indirect costs were also significantly associated with the number of previous disc surgeries: the more disc surgeries patients received in the past, the higher costs occurred. Having multiple disc surgeries indicates a worse health state, e.g. due to a chronic back pain disease with more severe spine involvement or more complicated surgery conditions. This may result in higher direct costs due to more treatments and higher indirect costs due to more sickness absence.
Our regression analysis showed significant associations of direct and indirect costs with gender. Female gender was associated with higher direct but lower indirect costs. A deeper view into direct costs showed that women had higher costs in almost all cost categories for lumbal and cervical disc herniations regardless of psychiatric comorbidity being present or not. One possible explanation for this finding may be that women in our sample were in worse health states. Women had more often comorbid chronic medical conditions (44% vs. 35%) and received on average more previous disc surgeries (0.29 vs. 0.19). Both variables were associated with higher direct costs in the regression analysis which may partially explain the gender effect. For indirect costs, we interpret the finding of lower costs in women - at least in part - as an artefact resulting from lower productivities applied for the monetary valuation of lost productivity time.
Interestingly, indirect costs were significantly (p
0.05) negatively associated with the number of children and private health insurance. We interpret these two findings as results of selection bias. On the one hand, one could assume that patients with children have more pressure to return to work, which may result in reduced indirect costs. On the other hand members of private health insurance tend to be healthier due to risk selection of private insurers. Further, members of private health insurance in Germany often earn higher income or are self-employed which both may be incentives to return to work fast.
In our base case analysis we excluded direct informal care costs and indirect costs resulting from reduced productivity at work. Costs of informal care were excluded because they are somewhat hypothetic: informal care costs present monetarily valued care time of relatives or friends [38
]. Thus – in contrast to all other direct costs – no “real” money is paid. Instead, informal care costs represent the opportunity costs of leisure time lost by relatives or friends. Productivity reduction at work was assessed by a ten step Likert scale ranging from “no reduction of productivity” to “unable to work” on which patients were asked to rate themselves. This scale has not been validated yet, therefore we excluded productivity reduction at work from our base case analysis too. Including these two cost categories in alternative analyses primarily resulted in clearly larger constants and larger effects of employment status and gender but no fundamental differences in results like sign changes.
Our statistical models explained a great share of the overall variance in costs, with coefficients of delimination (R2) ranging from 0.34 to 0.48. One must note that these high values of R2 are in part a result of pseudo-variance generated by the variables disc location and employment status. Whereas average costs of disc surgery were 3,572 Euro for lumbal disc herniation, they were 5,618 Euro for cervical disc herniation. Indirect costs predominantly occurred in employed patients resulting in average 3-month indirect costs of 5,338 Euro for those employed compared to 1,907 Euro for those unemployed. Thus the relative high values of R2 generated by our models should be seen with caution.
Our study has some limitations. We found relative high portions of missing values in DRG rates and the number of drug packages used. DRG coding required a complete set of variables, including hospital record data which were often not available, whereas the high portion of missings for drug packages may be due to memory effects. Our sample contained patients with cervical and lumbal disc herniations, which may bias our results. However, disease specific and sociodemographic characteristics were similar in both patient groups; furthermore we controlled for disc location in our regression analysis. Our assessment of psychiatric comorbidity was restricted to the most important CIDI sections (affective, anxiety and substance use disorders) that represent the most prevalent and costly psychiatric disorders. Further, the assessment of psychiatric comorbidity took place after the surgical intervention and may be influenced by this acute event, resulting in an overestimation of prevalence rates. Finally, some prices were not from our base year 2007, because no prices for this year were available. Instead we were forced to use prices of 2008 and 2009 for some goods. However, the portion of costs affected by this bias was very low und should not have a significant effect on the results.
Implications for clinical practice
Our findings imply that more attention should be given to psychiatric comorbidity in the back pain patients undergoing disc surgery. Clinicians should be aware of the high prevalence rates of psychiatric comorbidity in back pain patients, in particular in the most severe cases which are treated via surgery. If applicable, they should consider the assessment of psychiatric distress and support of mental health professionals [39
]. Multimodal diagnostic and therapy approaches that pay attention to psychiatric comorbidity may help to improve the outcomes of surgical therapy and to reduce the costs connected to psychiatric comorbidity.