OPBSResearch

Must Know

Real-world evidence (RWE) comes from routine care data such as EHRs, billing data, registries and cohorts.
The FDA wants RWE for postmarketing surveillance of implantable devices.
Registries and cohorts usually track implant survivorship, adverse events and patient-reported outcomes.
RCT results are hard to extrapolate because real patients and health systems vary more.
Registries and cohorts with diverse real-world patients narrow the gap between trial efficacy and practice.
EHRs serve care and billing but their data elements are rarely defined consistently across health systems.
Well-designed registries with consistent data are needed for comparative effectiveness research.
A learning health system shares data both ways. Care data feed research and registry results return to the EHR.
The guiding principle is collect once, use many times.

Registry Advantages Over EHR Data

Problem with EHR data	Registry solution
Irregular visit timing	Predefined measurement times across settings
Patients who stop attending clinic are lost	Automated prompts by web, phone or office to all patients
Missing documentation read as absent disease	Explicit recording of each comorbidity
Inconsistent measures, for example smoking	Core measure sets with one defined format such as pack years
No death or long-term endpoints	Linkage to national death indices and administrative data
Single site only	National or regional benchmarks
Weak statistics capability	Advanced, risk-adjusted analysis

Pooling many sites accumulates patients faster, so good and bad signals appear sooner.
APIs let EHRs send data to registries and receive risk-adjusted metrics in real time.
FORCE-TJR returns risk-adjusted predictions to patients and surgeons for shared decisions.

Landmark Cohorts and Trials

Study	Design	Main point
SPORT	Multicentre RCT with concurrent observational cohort, 13 US centres	Disc herniation ITT showed no difference in primary outcomes. As-treated analysis favoured surgery at 8 years
MOON	Prospective multicentre ACLR cohort, over 3,500 patients	Failure linked to age and graft type. Follow-up at 2, 6 and 10 years
MARS	Prospective revision ACLR cohort, 1,205 patients, 83 surgeons, 52 sites	Allograft raised early failure and lowered sports scores
MeTeOR	Multicentre RCT, 351 patients, 7 sites	Physiotherapy and partial meniscectomy equal at 6 and 12 months in ITT. Substantial crossover

SPORT had heavy nonadherence, so ITT and as-treated results differed.
MeTeOR supports physiotherapy first, with later arthroscopy if symptoms persist.

Arthroplasty Registries

Registry	Facts
FORCE-TJR (2010)	AHRQ funded representative US sample. Over 30,000 procedures in 28 states. National PROM norms. Real-time feedback
MARCQI (Michigan)	State consortium. Captures 97 percent of Michigan joint replacements. Over 250,000 cases
Kaiser Permanente	About 250,000 joint replacements since 2001. Implant revision risk and DVT prophylaxis studies
AJRR (2012)	Voluntary. About 28 percent of US joint replacement volume in 2017. Nearly 1 million procedures
SHAR, AOANJRR, NJR	Centralised systems capture most national joint replacements. Annual public reports

Annual reports give whole-registry, implant-specific and diagnosis-specific survivorship curves, mortality and resource use.
Registries can identify outlier implants that perform poorly.
Europe has over 20 arthroplasty registries holding over 5 million procedures.
National registries offer large samples with patient and clinician heterogeneity for strong RWE.

Exam Traps

EHRs were designed for care and reimbursement and cannot ensure consistent data across systems.
AJRR membership is voluntary and captures a minority of US volume.
Scandinavian, Australian and UK registries achieve high capture because healthcare is centralised.

Related pages. Bearing Surfaces and Wear, Level of Evidence

Recent Review. Large Database Research

Assess any database with the 5 Ws. Who is included, what is captured, where, when and why it exists.
Administrative databases come from billing codes (ICD-10, CPT). Examples are Medicare, NIS, PearlDiver and MarketScan.
Administrative data suit trends, geographic variation, short-term complications and cost analysis.
Clinical registries are treatment focused, hold procedural and implant detail, give survivorship and act as early warning systems.
Clinical registries and chart-abstracted data (NSQIP nurse reviewers) are more accurate than claims data.
NIS is inpatient only and NSQIP follows patients for just 30 days.
Large database studies give Level III or IV evidence and show association, never causation.
Two databases asked the same question often give different answers.
Case-control designs suit outcomes too rare for an RCT or cohort study.
CONSORT appraises RCTs. STROBE appraises cohort and case-control studies.

More on observational versus randomised evidence

An a priori research question avoids a fishing expedition through the data.
NSQIP laboratory variables can be 15 to 45 percent missing.
Coding changes such as the ICD-9 to ICD-10 switch can confound trend analyses.
RCTs control for unknown confounders through randomisation and blinding, but report average effects.
The PATH statement encourages reporting RCT results stratified by a risk model to show heterogeneity of treatment effect.
A Cochrane review found a pooled ratio of 1.08 (95 percent CI 0.96 to 1.22) between observational and RCT effect estimates.
Underpowered or poorly run RCTs are hypothesis generating and should not drive guidelines.

Browne JA, Springer B, Spindler KP. Optimizing use of large databases in joint arthroplasty and orthopaedics. J Bone Joint Surg Am. 2022;104(Suppl 3):28-32. PubMed

Recent Review. International Arthroplasty Registries

The Swedish Knee Arthroplasty Register (1975) was the first national quality register. The Swedish hip register followed in 1979.
Nordic personal identity numbers allow unambiguous linkage between national registers, with 90 to 99 percent completeness.
The UK NJR began in 2002 and held over 3 million procedures by 2020. Case ascertainment rose from 43 percent to 90 percent.
The AOANJRR began in 1999, reached national coverage in 2003 and has close to 100 percent voluntary surgeon participation.
AOANJRR status as a federal quality assurance activity protects data from subpoena and identification.
AOANJRR captures about 97 percent at first submission and validates against state health discharge data.
An outlier device has more than twice the class revision rate with a statistically higher hazard ratio.
The AOANJRR first flagged the ASR hip in 2007 and first showed lower revision with cross-linked polyethylene.
Registries record revision only. Symptomatic patients who are not revised are missed.
Registry data are observational, open to selection bias and residual confounding, and cannot prove causation.

Source

Franklin PD, Lange JK. Role of Cohort Studies and Registries in Musculoskeletal Healthcare and Research. In. Aaron RK, editor. Orthopaedic Basic Science. Foundations of Clinical Practice. 5th ed. Philadelphia, Wolters Kluwer, 2021.

Practise Type X questions on this topic

Reviewed by Professor Phong Tran, Head of Orthopaedic Surgery, Western Health. Last updated 17 October 2026.