Auditing Practices of the Medical Service Regarding Hospitalizations of Patients With a Secondary Diagnosis of Obesity

Abstract

Background

It has been observed that the Medical Service (Medizinischer Dienst, an auditing body of the German statutory health insurance system) is more likely to audit the bill for a hospitalization in a psychosomatic clinic if the patient carries a secondary diagnosis of obesity.

Methods

In an exploratory study, we retrospectively analyzed 771 datasets collected in 2019 as part of the standard documentation of acute psychosomatic hospitalizations.

Results

In 2019, the Medical Service audited bills for psychosomatic hospitalizations much more often in obese than in non-obese patients (odds ratio [OR] 2.499; 95% confidence interval [1.69; 3.69]). This was accounted for by a very high audit rate for patients with a secondary diagnosis of grade 3 obesity (OR = 3.972 [2.30; 6.86]). The audit categories „quality of coding” and “possible incorrect admission” were examined.

Conclusion

Treatments of markedly obese inpatients that incurred greater expenses presumably led to a higher hospitalization audit rate as an automatic consequence of the auditing algorithms used. An unintentional statistical discrimination arose from the unjustified linkage of the audit category “quality of coding” of the secondary diagnosis (obesity) with the audit category “possible incorrect admission” with regard to the main diagnosis. Similar effects may be occurring with economically relevant secondary diagnoses in other areas of medicine as well.

This study was prompted by the observation that the German Medical Service (Medizinischer Dienst, an auditing body of the German statutory health insurance system) is more likely to audit a bill for hospitalization in a psychosomatic clinic if the patient carries a secondary diagnosis of obesity compared to non-obese patients.

The basis for hospital billing in the area of psychosomatics and psychotherapy is the flat rate payment system in psychiatry and psychosomatics (PEPP) (1), which was introduced in 2009 in order to, among other things, achieve more performance-oriented remuneration. The PEPP remuneration system groups together various main diagnoses with similar levels of treatment cost into diagnosis related groups (DRGs), which are all remunerated in the same way. Additional characteristics such as age, receiving special services (procedures), as well as particular secondary diagnoses result in the case being put in a group that is more highly remunerated (referred to as upcoding).

The introduction of the PEPP remains a controversial subject (2). In addition to the criticism of the actual implementation of the PEPP, there are also doubts about its usefulness (3). Particular points of criticism include the bureaucratic consequences of the PEPP, whereby considerable documentation requirements are placed on hospitals, and the regular audits carried out by the Medical Service. Scientific insights into the effects of the flat-rate payments are very limited (4).

As has already been the case in the somatic sector, the introduction of a payment system based on DRGs caused a steady increase in the Medical Service’s auditing of hospital bills. To counteract this development, the German Medical Service Reform Act (MD-Reformgesetz) (5) that came into force at the end of 2019 limited the option to raise legal challenges, and audit quotas were set for bill audits. Due to the coronavirus pandemic, numerous regulations have been put on hold, and it remains unclear what effect the German Medical Service Reform Act will have. As part of the legal changes, a new audit procedure was agreed upon; however, the key points of bill auditing remained unchanged (6, 7).

Accordingly, health insurance funds are obliged to audit hospital bills with regard to anomalies suggestive of incorrect billing or uneconomical hospital treatment. If they identify anomalies, the health insurers cite one or more audit categories, and this may lead to a billing audit being conducted by the Medical Service.

A distinction is made between four different reasons for triggering an audit:

• Primary incorrect admission (Was inpatient treatment necessary in the first place?)

• Secondary incorrect admission (Could the patient not have been discharged earlier?)

• Coding audit (Were diagnoses and procedures correctly made/performed and coded?)

• Issues relating to the precondition for certain measures.

Patients with obesity represent a relevant group in terms of health care in Germany. According to data from the Robert Koch Institute, the most recent representative population survey of 18- to 79-year-olds revealed that 23.3% of males and 23.9% of females were obese (BMI ≥ 30) (8). In more detail, 2.8% of females and 1.2% of males (8, 9) had grade 3 obesity (body mass index [BMI] ≥ 40). It is well known that people with obesity are often stigmatized and discriminated against (10). This affects many areas of life of this group, including their health care (11, 12).

There are clear associations between obesity and the occurrence of mental illness. The risk, expressed in odds ratio (OR), for developing depression in obesity is between 1.2- and 5.8-fold higher compared to normal-weight individuals (13). Of particular importance in terms of the psychological comorbidity of obesity is binge eating disorder (BED), an eating disorder that involves episodes of overeating and which occurs in approximately one-third of obese patients (14). Therefore, an increased proportion of comorbid obesity is to be expected in the hospital treatment of patients with mental illness.

For a variety of reasons, there is also greater use of resources in the inpatient treatment of obese patients (technical facilities, special treatment options, increased care requirements). Grade 3 obesity is one of the complicating secondary diagnoses that generates a higher flat-rate remuneration (upcoding from PP04B to PP04A) in the psychosomatic treatment of affective, neurotic, stress, somatoform, and sleep disorders. The prerequisite for this, however, is that the secondary diagnosis is correctly coded in line with the German coding guidelines (15): According to this, secondary diagnoses are interpreted as disorders that affect patient management in such a way that one of the following factors becomes necessary:

• Treatment measures

• Diagnostic measures

• Greater requirements in terms of care, nursing, and/or monitoring

To check for anomalies, the health insurance funds have at their disposal digital datasets of hospital bills and services, which are automatically analyzed digitally using billing programs. Which algorithms are used for this is not known to us.

However, the use of algorithms harbors the risk of discrimination, as shown in a study by Orwat (16). Discrimination is often seen as disadvantaging and unjustified unequal treatment of individuals with a protected characteristic, for example, a particular health condition. If there is an objective reason for the unequal treatment, it may be socially acceptable. If one uses algorithms that result in unequal treatment that cannot be objectively justified, this is referred to as statistical discrimination. This can be unintentional or hidden.

Against this backdrop, the present exploratory study aims to address the following questions:

• Does the diagnosis of obesity in patients hospitalized for psychosomatic disorders increase the likelihood of an audit by the Medical Service?

• What role do possible variables play?

• Does the grade of obesity (grade 1–3) affect the likelihood of an audit by the Medical Service?

• What effect does the diagnosis of obesity have on the distribution of reasons to carry out an audit and the rate of objections raised by the Medical Service regarding hospital bills?

Methods

The present retrospective study is a secondary analysis of pseudonymized data from a hospital for psychosomatics and psychotherapy in Hesse, Germany, focusing on the treatment of general psychosomatic disorders, depressive disorders, post-traumatic disorders, and eating disorders. The data source was the standard documentation of the Parkland Clinic in Bad Wildungen for the last calendar year prior to the pandemic.

The study was carried out in accordance with the provisions of the Declaration of Helsinki; the responsible ethics committee raised no objections (2021–2440-evBO). The dataset was pseudoanonymized prior to analysis using SPSS 21 software and did not leave the institution.

A total of 771 datasets of an original 925 patients discharged during the study period were retrospectively analyzed. In all, 154 cases were excluded due to lack of consent to use data for research purposes.

Descriptive statistics (frequencies, cross-tabulations, comparisons of means) and logistic regression analyses (method: inclusion) were used for the analysis. The following parameters were considered as independent variables:

• Secondary diagnoses of obesity and BED

• Other economically relevant diagnoses (bipolar disorder, schizophrenia, uncontrolled diabetes)

• The basic sociodemographic variables age and sex

• The number of treatment days as factors presumed to influence the dependent variable “audit by the Medical Service.”

The Wald test was used to assess whether the included predictors had a significant effect. To estimate the strength of the effect, ORs were calculated.

The diagnosis of BED, which is often comorbid with obesity, was included—as were age (up to 64 versus 65 and older) and sex—in subgroup analyses by stratifying the logistic regression analyses.

However, due the exploratory nature of the study, the entire data analysis ultimately remains descriptive Therefore, any statistical correlations found in the analysis are to be seen as evidence on which to generate, not to confirm, hypotheses.

Results

Of 771 analyzed cases, 194 cases (25.2%) were audited by the Medical Service. Table 1 provides a summary of patient characteristics.

Table 1. Sociodemographic characteristics, diagnoses, and length of hospital stay of the studied patient collective.

Characteristic	Value
Sex
Female	613 (80%)
Male	158 (20%)
Diverse	0 (0%)
Age
Age; mean (SD)	43.7 (16.1)
Age≥ 65 years	49 (6%)
Main mental health diagnosis according to ICD-10
Affective disorders	386 (50%)
Anxiety, stress, and somatoform disorders	200 (26%)
Eating disorders	140 (18%)
Personality and behavioral disorders	35 (5%)
Other mental disorders	10 (1%)
Economically relevant secondary diagnoses
Grade 3 obesity (BMI ≥ 40)	59 (8%)
Bipolar disorder	8 (1%)
Schizophrenia	2 (0%)
Decompensated diabetes	0 (0%)
Length of hospital stay
Treatment days; mean (SD)	52.1 (23.4)

SD, standard deviation; BMI, body mass index

In order to determine the extent to which the potential factors obesity, BED, treatment days, sex, and age contributed to an audit by the Medical Service, a logistic regression was calculated. Three further possible factors eluded viable analysis due to insufficient case numbers (bipolar disorder, schizophrenia, decompensated diabetes; n < 10 in each case) (17).

The regression model as a whole was significant (p < 0.001). A total of 74.8% of cases could be correctly classified. The diagnosis of obesity increases the likelihood of an audit by the Medical Service, as does a longer treatment duration (per day; Table 2).

Table 2. Logistic regression analysis for the prediction of a Medical Service audit.

Predictor	Odds ratio [95% CI]	Significance*
Sex	0.981 [0.636; 1.512]	0.930
Age ≥ 65 years	0.765 [0.357; 1.638]	0.490
Obesity	2.499 [1.690; 3.694]	< 0.001
BED	0.947 [0.437; 2.048]	0.889
Treatment days	1.024 [1.016; 1.032]	< 0.001

* Wald test

BED, binge eating disorder, CI, confidence interval\

Table 3 shows the effect of the predictor obesity. The billing data of patients with a diagnosis of obesity (n = 170) were significantly more frequently audited by the Medical Service compared to patients without a diagnosis of obesity (37.1% versus 21.8%).

Table 3. Diagnosis of obesity and MS audit.

Diagnosis of obesity	Audit by the MS	No audit	Total
Yes	63 (37.1%)	107 (62.9%)	170 (100%)
No	131 (21.8%)	470 (78.2%)	601 (100%)
Total	194 (25.2%)	577 (74.8%)	771 (100%)

MS, Medical Service

As can be seen in Table 4, the effect of the predictor obesity was largely independent of sex, but not of age group of the patients: Only in younger patients (under 65 years) did the presence of obesity emerge as a risk factor for an audit by the Medical Service. In the case of patients with BED in addition to obesity, the likelihood of a Medical Service audit was twice as high compared to obese patients without BED. However, the corresponding ORs in the small groups of older patients and patients with BED were based on non-significant regression models (omnibus tests: p = 0.996 and p = 0.159, respectively) and therefore should be interpreted with caution.

Table 4. The risk obesity has on case audit by MS, differentiated by age, sex, and presence of BED.

Analysis group	Odds ratio [95% CI]	Significance*	N
Total collective	2.499 [1.690; 3.694]	< 0.001	771
Sex
Male	2.773 [1.285; 5.987]	0.009	158
Female	2.401 [1.544; 3.732]	< 0.001	613
Diverse	–		0
Age
Up to 64 years	2.632 [1.777; 3.896]	< 0.001	722
From 65 years	0.984 [0.173; 5.596]	0.986	49
BED
No	2.427 [1.623; 3.628]	< 0.001	734
Yes	4.255 [0.746; 24.257]	0.103	37

* Wald test

BED, binge eating disorder; CI, confidence interval; MS, Medical Service; OR, odds ratio

When considering the different grades of obesity, it was found that hospitalizations of patients with grade 1 or grade 2 obesity were not audited more frequently by the Medical Service than were those of non-obese patients. The higher audit frequency found for the group of obese patients is based on a very high audit rate for patients with grade 3 obesity (OR 3.972; 95% confidence interval: [2.300; 6.860]). In all, 31 of 59 patients (52.5%) with grade 3 obesity were audited, but only 163 of 712 patients (22.9%) without this diagnosis.

None of the Medical Service audits of patients with a diagnosis of obesity (grades 1–3) reviewed quality of coding alone (table 5). Primary as well as secondary incorrect admission were audited significantly more frequently in the case of hospitalization of patients with obesity (OR 1.832; [1.197; 2.804] and OR 2.142; [1.476; 3.107], respectively) compared to patients without a diagnosis of obesity.

Table 5. Audit categories.

Audit category	Diagnosis of obesity	No diagnosis of obesity	Total
Coding only	0 (0%)	2 (1.5%)	2 (1.0%)
Coding audit and primary incorrect admission	0 (0%)	1 (0.8%)	1 (0.5%)
Coding audit and secondary incorrect admission	10 (15.9%)	19 (14.5%)	29 (14.9%)
Primary incorrect admission only	3 (4.8%)	6 (4.6%)	9 (4.6%)
Secondary incorrect admission only	14 (22.2%)	26 (19.8%)	40 (20.6%)
Primary and secondary incorrect admission	18 (28.6%)	57 (43.5%)	75 (38.7%)
All three audit reasons	18 (28.6%)	20 (15.3%)	38 (19.6%)
Total	63 (100%)	131 (100%)	194 (100%)

With 14.3% objections to hospital bills in the initial Medical Service audit, the objection rate for bills with a secondary diagnosis of obesity was almost twice as high as for bills without this diagnosis (7.6%). If one makes a differentiation on the basis of audit category, one clearly sees that objections were primarily raised regarding “quality of coding.” In contrast, primary or secondary incorrect admissions were found approximately as often in the patient group with an obesity diagnosis as in the group without.

Discussion

The results of the present study suggest that 2019 bills for psychosomatic hospitalizations of obese patients were audited by the Medical Service significantly more frequently than were those of non-obese patients. This applies not only to the audit category “quality of coding” but also to the category “possible incorrect admission.” The higher audit frequency for the group of obese patients is based exclusively on a very high audit rate among patients with grade 3 obesity.

When examining possible variables, a surprising effect is clearly seen for age; however, this requires statistical validation. From the age of 65 onwards, hospital bills of obese patients are not audited more frequently than those of non-obese patients. On the contrary, older age appears to have had some form of “protective function” against a Medical Service audit.

An evident reason for these observations is different remuneration of the treatment: Age over 65 years and grade 3 obesity are subject to the same (higher) flat-rate reimbursement. However, since age cannot be called into question, a Medical Service audit of an obesity diagnosis makes no sense in this case, since it would need to be billed at a higher flat rate anyway. The higher flat-rate remuneration is presumably the reason for the very high audit rate for secondary diagnoses of grade 3 obesity. Grade 3 obesity results in higher flat-rate remuneration in the PEPP—grade 1 and 2 obesity do not.

This association is suggestive of the use of audit algorithms that are designed for cost optimization, not primarily for the characteristic obesity (grades 1–3).

Thus, the identification of a more highly reimbursed diagnosis of grade 3 obesity is the starting point for a Medical Service audit. Whether or not the mere existence of a diagnosis that is reimbursed at a higher rate constitutes an objective reason to suspect a coding error and hence trigger a Medical Service audit is, in our view, debatable. However, the truth of the matter is that this auditing practice is accepted by those involved as objectively justified.

In our study, an audit of coding quality of the secondary diagnosis grade 3 obesity was associated in all cases with an audit of primary or secondary “possible incorrect admission,” that is to say, an audit of the indication for treatment based on the principal diagnosis. At this level, it is probable that the audit algorithm used by the health insurance funds creates an unobjective linkage between the characteristic “grade 3 obesity” and potentially uneconomical treatment of the main diagnosis (“possible incorrect admission”).

In practice, this means: If a patient is treated for a depressive disorder in a psychosomatic hospital, they are four times more likely to have their stay audited for suspected uneconomical treatment if they have grade 3 obesity as a secondary diagnosis.

Thus, we consider it is possible that the fact that there is a higher rate of audits may influence decisions regarding the admission and length of treatment of obese patients in a psychosomatic clinic. For the discrimination criteria according to Orwat (16) to be met, it is not necessary to prove a direct disadvantage. The unobjective linkage of “quality of coding” and “possible incorrect admission” with the secondary diagnosis of grade 3 obesity in the auditing algorithm used for hospital bills indicates unintentional and previously hidden statistical discrimination.

The high rate of objections raised by the Medical Service when auditing the coding quality of the diagnosis ‘obesity’ needs to be viewed in a differentiated manner. It suggests only in part that there were indeed coding deficiencies. In some cases, the final court decisions are still pending. It is possible that the difficulty in correctly representing comorbid BED also plays a role in the coding shortcomings identified. Since ICD-10 does not include BED, three different codes for BED are given side by side in clinical practice: F50.4, F50.8, and F50.9. Moreover, BED is challenging to diagnose (18). The increased rate of audits for the primary and secondary incorrect admission of patients with a secondary diagnosis of obesity does not result in a higher rate of challenges by the relevant Medical Service. Thus, at the Medical Service level, there is no evidence of discrimination.

Limitations and strengths

There are weaknesses in the study design: It is a purely exploratory monocentric study. Particular features of the study center may have led to a higher rate of audits of one diagnosis group. Having said, the comparison group also comes from the study center. Regional effects in the evaluation of launched audits are unlikely. Patients are assigned to care nationwide, and billing audits are carried out centrally by the major health insurers. A consideration of the audit results of only one regional Medical Service has modest informative value. The extent to which the results can be extrapolated is limited by the as yet unclear consequences of the German Medical Service Reform Act.

The study’s strengths include the statistical validation of the exploratory investigation and, in the research context, the novel question being posed.

Summary

In this work, we have demonstrated that higher reimbursement for the inpatient care of severely obese patients as a result of the use of audit algorithms leads to a higher rate of audits of the entire hospital stay. In this context, we consider the occurrence of diagnosis-related discrimination to be inherent in the system: In order to function, a DRG system requires diagnosis-related audits. These only work with audit algorithms and, in certain cases, can result in unintentional discrimination.

There is good reason to believe that likewise in other specialties, hospitalizations with economically relevant secondary diagnoses are more likely to be audited. The possibility of statistical discrimination in the DRG system should be evaluated in studies with a broader data basis. Due to the sensitive nature of the data, a research undertaking of this kind is only possible with the support of the legislator and the self-governing bodies. A discussion regarding possible solutions could then be had on a solid basis.