J Manipulative Physiol Ther 2004 (Sep); 27 (7): 442–448 ~ FULL TEXT
Shawn P. Phelan, DC, Richard C. Armstrong, DC, David G. Knox, DC,
Michael J. Hubka, DC, Dennis A. Ainbinder, MD
OBJECTIVE: To examine utilization, treatment costs, lost workdays, and compensation paid workers with musculoskeletal injuries treated by medical doctors (MDs) and doctors of chiropractic (DCs).
DESIGN: Retrospective review of 96,627 claims between 1975 and 1994.
RESULTS: Average cost of treatment, hospitalization, and compensation payments were higher for patients treated by MDs than for patients treated by DCs. Average number of lost workdays for patients treated by MDs was higher than for those treated by DCs. Combined care patients generated higher costs than patients treated by MDs or DCs alone.
CONCLUSIONS: These data, with the acknowledged limitations of an insurance database, indicate lower treatment costs, less workdays lost, lower compensation payments, and lower utilization of ancillary medical services for patients treated by DCs. Despite the lower cost of chiropractic management, the use of chiropractic services in North Carolina appears very low.\
From the Full-Text Article:
This study captures a data set from a large population of injured workers (96,627) over a relatively long period of time (19 years). The inclusion criteria retained 43,650 claims and encompassed all cost aspects of treatment. In addition to the variables discussed in the Methods section, information was captured on injury type, prevalence, and frequency of presentation for both nature of injury and body part. Additionally, the study defines specific provider utilization for the 18 separate categories of musculoskeletal conditions analyzed. The information was extracted by information technology (IT) specialists for the State of North Carolina under the auspices of the Department of Commerce. The IT specialist who designed the original archival system in 1974 designed and conducted the data extraction for this study. Biostatisticians jointly at the University of California (UCLA) and the Southern California University of Health Sciences (SCUHS) performed the analysis. The authors did not assist with either the extraction or the analysis of the data.
Although these data contain all reported injuries archived during the available 19-year time frame, there is variability to the information contained in the individual files. This may be because of changes in data entry policy, technology, or even staff. Additionally, the data captured may not contain all treated work-related injuries that occurred during this time. Treatments can be rendered without a report of injury to the North Carolina Industrial Commission. It is not the responsibility of the MD or the DC to report the injury. It is instead the responsibility of the employer and/or carrier. This eliminates the possibility of provider reporting bias for the purposes of this study. Without a report to the North Carolina Industrial Commission, the injury would not appear in the study data. Also, there were 52,997 closed claims that contained only compensation payments, because provider type had been purged. These claims were dropped from the study but would have contained additional patients for both provider types. These additional claims may have impacted the findings.
Inherent inaccuracies in insurance databases present challenges of methodology. These challenges include lack or inaccuracy of diagnostic and procedural codes, lack of control for acuity and chronicity, incomparability of groups, absence of information on prognostic indicators, insufficient outcome measures, lack of information on comorbidity, and missing data. [7, 11, 19, 20, 21] However, these databases are representative of defined populations treated within a defined fee schedule without provider bias. This results in having an excellent ability to generalize the results to a wider population. Despite the limitations, insurance databases provide a stable frame of events and remain a legitimate source for the study of utilization. [22, 23]
This study did not elude the methodologic difficulties of previous studies. The most notable limitations are the lack of data on severity of injury and comorbidity. Management costs of injured workers in North Carolina during the years 1975 to 1994 were defined with this study. However, we cannot determine if either provider group treated the more severe or chronic musculoskeletal injuries. If inpatient and outpatient hospital costs are an indicator, then the MDs would seem to have treated the more severe and, hence, the more costly injuries. Unfortunately, this cannot be determined with the available data. Accessing the physical files of this database and manually extracting the information could show additional diagnostic information. Severity, comorbidity, acuity and chronicity, as well as patient's age and sex could then be differentiated. Prospective studies and randomized trials are needed to continue the investigation as well. 
The data in the state of North Carolina provide a picture, though incomplete, of utilization rates and management costs of injured workers. Given the results of this study, utilization of DCs for the treatment of injured workers is very low. There are no legislative impediments to injured workers wishing to access DCs, and the North Carolina Industrial Commission Medical Fee Schedule allows for full scope of practice reimbursement for DCs. A survey of North Carolina patients in the general population suffering acute, severe low back pain found 13% of study respondents initially sought care with a chiropractor.  Although a comparison of this population's utilization rates to workers' compensation rates may be inappropriate, it raises questions. A review of a Liberty Mutual Insurance nationwide workers' compensation databank showed a chiropractic utilization rate of 3%.  Comparing these rates with the 0.8% utilization rate of chiropractors in our North Carolina data may suggest that barriers to injured worker access to chiropractors exist in North Carolina.
The differences in provider management costs, independent of critical issues such as severity and comorbidity, suggested by these results indicate lower treatment costs, fewer lost workdays, reduced utilization of ancillary medical services, and reduced compensation payments for patients treated by DCs. Recognizing the study limitations, if indeed the provider subsets are comparable, it seems likely that substantial savings to the workers' compensation system would be possible if chiropractic services were increased in North Carolina.
Combined care patients tended to have significantly higher costs across all categories. Unfortunately, these data do not allow us to determine why their costs were so much higher. It is possible that factors such as symptom chronicity, initial provider selection, potential litigation, or patient satisfaction caused these cost increases.