Minnesota Workers’ Compensation Program: Mandatory Treatment Parameters Evaluation The Robert Wood Johnson Foundation Workers’ Compensation

Health Initiative
David Gilbertson, Ph.D. Stratis Health William Lohman, M.D. Minnesota Department of Labor & Industry

Goals for the Evaluation
• To ascertain provider compliance with the Treatment Parameters for low back pain. • To investigate the impact of the Treatment Parameters for low back pain on quality and outcomes of care. • To identify opportunities for improvement in the workers’ compensation mandatory medical treatment parameters for low back pain.

Model for the Evaluation
Treatment Claimant Outcomes Clinical Environment

Evaluation Design
• Longitudinal follow-up • Inception cohort • Repeated cross-sectional surveys • Supplemented by administrative data • Partitioning on adherence to parameters

Worker is injured 1. Employee sees a medical provider

2. Provider reports injury to Insurance Carrier

Insurance Carrier
Medical bills Medical record

9. Abstract record information from insurance carrier after 6 months from DOI

3. Insurance Carrier reports injury to DOLI

DOLI
Administrative Database
DOI Type of injury Demographics Employment Return to work Indemnity cost Litigation

6. Stratis Health sends copies of consent forms and list of Patient I.D.’s

Stratis Health
Consent to participate Baseline survey 6 Month survey

7. DOLI provides name, address, telephone #, and DOI

4. DOLI sends baseline survey and “consent to participate”

5. Claimant mails baseline survey and “consent to participate”

Claimant
Demographic Health status Employer & provider satisfaction Treatment

8. Stratis Health sends 6 month follow-up survey

Sources of Data
• Initial Survey within one month after eligible First Report of Injury received by DLI • Six-Month Follow-Up of respondents to Initial Survey • Medical Bills and Records of respondents to Initial Survey • Administrative Data for all eligible participants

Initial Survey
• Started in April, 1997 • Closed enrollment on December 31, 1997 • Single mailing from DLI • Eligibility rate = 20.1%

Initial Survey
• Case ascertainment • Low back pain and functioning • Severity classification • Change in health, pain, functioning, and daily activities since First Report of Injury

Follow-Up Survey
• Started in November, 1997 • First mailing, telephone reminder, second mailing • Response rate = 65.3% • No forwarding address for 2.7%

Follow-up Survey
• Claimant characteristics • Claimant outcomes • Modifying factors
– health behaviors – employment related – psychological

• Clinical characteristics

Follow-up Survey
Claimant Characteristics • • • • • Gender Age Marital Status Education Ethnic Background

Follow-up Survey
Claimant Outcomes • Functioning and well-being • Low back-specific measures • Return to work, disability, resources, and expenditures • Satisfaction with outcomes, care, and the quality of services

Follow-up Survey
Modifying Factors • • • • • Health Transitions Employment Satisfaction Smoking History Height & Weight Co-Morbidities

Follow-up Survey
Clinical Characteristics • • • • Active Modalities Durable Medical Equipment Medications Job Accommodations

Administrative Data
• • • • Indemnity benefit information Claim status Litigation history Claimant information
industry job tenure prior claims occupation wages

Administrative Data
Administrative Outcomes

• Benefits paid • Duration of benefits

Medical Bills and Records
• Timing and frequency of health care services • Justification of departures from parameters

Data Sources
Initial Survey Analysis File
Insurance Carrier: Medical Bills and Records MN Dept. of Labor & Industry: Administrative Data

Follow-up Survey

Claimant Cohorts
• Potential eligibles with FRI during 1997 with one of the included injury codes (FRI Cohort) • Claimants meeting the case definition and agreeing to participate in the project (Recruitment Cohort) • Members of the Recruitment Cohort who returned a completed follow-up survey (Follow-up Cohort) • Members of the Recruitment Cohort for whom medical bills and records were obtained (Analysis Cohort)

Relationship Between Claimant Cohorts
FRI Cohort True Eligibles Recruitment Cohort Follow-up Cohort Analysis Cohort

Claimant Cohorts
F R I C o h o rt 5074 R e c r u it m e n t C o h o r t 1007 F o llo w - U p C o h o r t 638 A n a ly s is C o h o r t 626

Measuring Severity
• Type of Low Back Pain
• regional vs. radicular vs. “mixed”

• • • •

Quantitative Pain Drawing Pain Scales Initial Functional Limitation Prior History of Low Back Pain
• prior episode vs. prior treatment vs. prior surgery

• Patient Perceptions

Type of LBP:Regional Low Back Pain

Type of LBP:Radicular Low Back Pain

Type of LBP: “Mixed”

Measuring Severity
Type of Low Back Pain regional radicular 395 (61.9%) 243 (38.1%)

Measuring Severity
Quantitative Pain Drawing
pain paresthesia (12.2) front (112) 4.8 (9.5) 4.0 (10.1) back (112) 13.2 (11.8) 8.0

Measuring Severity
Pain Scales
at first visit… in last 7 days… now… Initial Functional Limitation Roland-Morris 81.8 (18.5) 39.3 (25.9) 31.5 (25.9)

10.5 (6.8)

Low Back Pain as Measured by Visual Analog Scale (VAS) from First Report of Injury to Recruitment Survey According to Low Back Classification

100 80

Mixed Radicular Regional

Pain Level

60 40 20 0

First Report of Injury

Last 7 Days on Average

Now

Measuring Severity
Prior History
previous LBP
• required treatment (93.7%) • had surgery (15.1%)

239 (37.5%)
224 36

Measuring Severity
Patient Perception -OVERALL HEALTH
excellent very good good fair poor

Measuring Severity
Transitions in Health States Since the First Visit
100 80 Percent (%) 60 40 20 0 Overall Health Activity Limitation Back Pain Worse Same Better

Measuring Severity
Patient Perception disabled permanently 204 (31.9%) 105 (51.5%)

Measuring Compliance
• Six different indices of compliance were developed and used during the analysis.
1 2 3 4 Use of provider focus groups to weight importance of parameters Use of “strength of evidence” scores from the AHCPR guidelines Dichotomous: overall compliance Dichotomous: within each of three components:
active treatments, passive treatments, diagnostic procedures

5 Count of non-compliant treatments/procedures: overall 6 Count of non-compliant treatments/procedures: within each of three components:
active treatments, passive treatments, diagnostic procedures

Measuring Compliance
• The a priori compliance index was based on provider opinions of the relative importance of the various components of the parameters in influencing the outcome of treatment. • Focus groups and repeated surveys were used to generate consensus in these opinions, and then weights reflecting this consensus were developed for each component of the parameters.

Measuring Compliance
Data for measuring compliance was obtained from:
– Medical bills in insurance records – Claimant reports in the Follow-up Survey

Measuring Compliance
A computer algorithm was developed to analyze compliance with treatment parameters:

LBP Treatment Parameters
flowcharts & timelines CPT codes & criteria

Computer Search Routine

Measuring Compliance
• Computer algorithm search of each claimant’s billing and questionnaire data to identify instances of possible noncompliance. • Cases with possible non-compliance flagged for review by MD, with final determinations after review of the claimant’s medical records.

Measuring Compliance
Computer search through HCFA 1500s

Compliant?

Unsure

Determination by MD review

Compliant?

Yes

No

No

Yes

Score assigned, aggregated within claimant

Analysis
• Each instance of non-compliance was weighted when aggregating into a final index of compliance for each claim. • Models were built for each outcome examined using demographic information, severity measures, prognostic factors, and claims characteristics.

Analysis
• After these models had been refined, the compliance score was added to determine the effect of compliance with the parameter on each outcome.

Results
• Average age 41 years (range:19 to 64 years) • Predominantly male (65.8%) • Most are married (65.8%) • At least high school graduate (93.2%) • White, not Hispanic (93.2%)

Results
• Overall Health
excellent very good good fair poor

slight decline in overall health

Results
90 80 70 60 50 40 30 20 10 0

percent (%)

worse same better

Overall Health

Activity Limitations

Back Pain

Results
Pain Scales
in last 7 days… now… Functional Limitation Roland-Morris improvement 30.5 (26.6) 27.7 (27.0)

7.3 (6.7) 3.2 (5.7)

Results
• Work Status
full-time part-time off work 463 (72.6%) 94 (14.8%) 80 (12.6%)

Results
• Off Work
back injury 52 another work injury 9 another medical problem laid off 16 personal reasons

5 16

Results
• Return to Work
same employer same job

55 7
487 (87.6%) 439 (79.0%)

• Accommodations
duties changed special equipment change in hours change in lifting

14 0
89 15 34 107

Satisfaction with Care
Courtesy

Explanation

Access

excellent very good good fair poor

Options 0 10 20 30 40

Percent (%)

Satisfaction with Care
Quality

Concern

Skill

excellent very good good fair poor

Control 0 10 20 30 40

Percent (%)

Satisfaction with Care
50 45 40 35 percent (%) 30 25 20 15 10 5 0 REFERRALS ALL Rx INVOLVED never sometimes usually always N/A

Results
• Of 626 claimants with complete information available, care delivered for 443 (70.8%) was completely compliant.
29% Compliant Non-compliant 71%

Results
Does Compliance Affect Outcomes?
improvement in pain improvement in function satisfaction with care satisfaction with job work status at 6 months mental health physical health lost work time no no no no no no maybe yes

Results
• In the physical component score of the SF-12 model, non-compliance on passive treatments was associated with decreased physical component scores. For each one-standard deviation increase in non-compliance, there was an associated 0.75 decrease in the physical component score.

Results
• In the return-to-work model, the six methods of computing compliance scores generally agreed in terms of effect on return to work of a one standard deviation change in compliance score. • For each standard deviation increase in noncompliance, the associated relative risk estimate from the proportional hazards models was 0.85.

Time to Return-to-Work for Compliant vs. Non-compliant Treatment
100 Percent not returned to work 60 80

Compliant Treatment Non-compliant Treatment

p < .001

0 0

20

40

5

10

15

20

25

# of Weeks (Total+Partial) Off Work

Time to Return-to-Work for Compliant vs. Non-compliant Treatment
After Adjustment for Severity
100 Percent not returned to work 60 80

Compliant Treatment Non-compliant Treatment 40

p < .01

0 0

20

5

10

15

20

25

# of Weeks (Total+Partial) Off Work

Results
• Return-to-work models investigating compliance within individual components (active, passive, and diagnostic) found increased compliance within passive and diagnostic treatments was associated with decreased time away from work.

Conclusions
• Adherence to the treatment parameters was associated with decreased duration of disability... … and produced equivalent results to less restricted treatment on the other outcomes assessed.

Conclusions
• A process combining administrative, medical utilization, and patient-reported data can be successfully implemented to evaluate the overall effects of a complex “episode of care” treatment guideline that integrates recommendations on diagnosis, treatment, and follow-up of cases over time.