Abstract
Background: PD was associated with increased mortality before levodopa therapy became available. There have been conflicting reports of PD mortality in the modern era.
Objective: To assess current mortality rates in a large unselected population receiving treatment for parkinsonism (PKM) followed for up to 6 years.
Methods: Cases were identified using linked administrative databases, including physician service and prescription drug claims, generated in Ontario’s universal health insurance system. Control subjects were identified from the provincial registry of citizens and age and sex matched to cases. Comparative mortality was evaluated over the 6-year period of the study (1993/94 to 1998/99). The sensitivity of the findings was tested with differing case definitions.
Results: In 1993, 15,304 patients with PKM were identified and were age and sex matched to 30,608 control subjects (1:2 ratio). Over the study period, 50.8% (7,779) of the cases with PKM died compared with 29.1% (8,899) of the control subjects. The cases with PKM had an overall mortality odds ratio of 2.5 (95% CI: 2.4, 2.6) compared with the control group. Results were consistent whether cases were defined by physician diagnosis, use of anti-PD drugs, or both criteria.
Conclusion: Despite modern drug therapy, PKM continues to confer a sharply increased mortality on unselected patients followed for several years.
PD is a common neurodegenerative disorder that has a considerable impact on the quality of life of persons with the condition.1,2⇓ Patients with PD and other forms of parkinsonism (PKM) have progressive disability, but the natural history of this disorder may be quite variable.3 PKM is associated with an increasing burden to society with large health care-related costs (M. Guttman, unpublished data).
A number of studies have attempted to estimate the associated mortality with PD, with inconsistent results reported. There are reports of increased mortality before current drug treatment became available,4 increased mortality due to selegiline,5 normalized mortality with levodopa therapy,6 improved mortality with levodopa that was maximal when treatment was initiated in the early stages of the condition,7 higher mortality despite the use of current antiparkinsonian drug treatment,8-15⇓⇓⇓⇓⇓⇓⇓ and lack of increased mortality in a large prospective study of early PD.16 These studies have many potential sources of error including small sample size, selection bias (since cases were referred to a tertiary clinic or included in a clinical trial), limited observation time to determine mortality, lack of a control group, use of death certificate data to make a diagnosis of PD, variable access of patients to health services, and inclusion of patients based solely on drug utilization.
In this study, we compared the mortality associated with PKM in an unselected population-based cohort as compared with age- and sex-matched control subjects. Since we were unable to differentiate patients with PD from those with other forms of PKM using the methodology employed, we refer to our findings in relationship to patients with PKM. Compared with previous studies, this study has the advantage of prospectively following and comparing a very large number of patients with PKM from the general population—identified by physician diagnosis and drug utilization—with age- and sex-matched controls within a universal health care system with accurate ascertainment of vital status.
Methods.
Setting/data sources.
The province of Ontario contains 38% of the Canadian population, with over 11 million residents. All Ontarians are entitled to physician services and all medically necessary hospital services through a publicly funded plan, with no direct out-of-pocket charges and no parallel private provision of these services. Those 65 years old and over and those on social assistance are also entitled to outpatient drug coverage. Administrative data generated by the health insurance system allow the examination of patient demographics, claims for physician services (the predominant payment mode is by fee per item of service), claims for outpatient prescription drugs in the population 65 years and over, and information about hospitalizations. Under our research agreements, all administrative data are stripped of personal identifiers and assigned encrypted unique numbers. This rendering of the data as anonymous protects the privacy of individuals yet allows complex linkage. The researchers do not have the ability to identify patients; it is therefore impossible to validate the administrative data by direct clinical examination of individuals.
For purposes of this study, we linked data on physician services, drug benefits for senior citizens, and vital status. The physician service data include the primary diagnosis made by the physician in relation to the service for which payment was claimed. The drug benefit data include claims for payment submitted by pharmacists to the government. Vital status was assessed using a registration file that includes the citizen’s health insurance plan number, sex, date of birth, postal code, and death date (where applicable).
Cases were identified by the use of the International Classification of Diseases, 9th revision, (ICD) diagnostic code for PD (332) recorded by the physician in conjunction with an outpatient service, by the use of specific PD drugs (defined below) in the drug claims database, or a combination of both. This technique was used to compile as complete a dataset as possible because physicians code only one primary diagnosis per visit. Patients were included if they were over the age of 25 during the inception year (fiscal 1993/94) on the assumption that a diagnosis of PD under this age represented a coding error.
Individuals had to meet at least one of the following inclusion criteria: at least one physician claim with ICD code 332 by a study physician (defined below), at least one claim for a drug used to treat PD (defined below), or a combination of at least one physician claim and at least one drug claim.
Although our methods were designed to ensure comprehensiveness, we were concerned that some physician claims with a primary diagnosis of PD might occur through clerical error or misdiagnosis. We therefore compiled a list of study physicians who were most likely to take an active role in the treatment of patients with PD and nonstudy physicians who were unlikely to be involved in the routine treatment of PD was compiled (see the appendix). Patients with a 332 code who were seen only by nonstudy physicians were excluded on the assumption that diagnostic coding errors would be more common in this group.
As a second check on miscoding or misdiagnosis, we also applied a temporal exclusion criterion. These administrative data first became available for research purposes for 1992/93. Patient information was therefore accumulated for the 1993/94 Ontario government fiscal year. To validate the inception cohort, data from the previous year (1992/93) were examined, and only those patients in the cohort who had physician claims and/or drug claims in 1992/93 and 1993/94 were included. If patients had claims in only 1993/94 and not 1992/93, they were excluded from the cohort. The cohort was then followed for 6 years (1993/94 to 1998/99) to obtain longitudinal data.
Specific drugs used to treat PD are rarely used for other conditions. Therefore, prescription drug information was used to help define or validate the diagnosis of PD. These drugs were included in the analysis, in both brand name and generic formulations as applicable: levodopa, levodopa/carbidopa (brand name, generic, and controlled release), levodopa/benserazide combinations, selegiline (brand name and generic), bromocriptine (brand name and generic), pergolide, tolcapone, ropinirole, pramipexole, and amantadine. Patients who were treated solely with amantadine during the influenza season (between November and April) were excluded. As the drug benefit database does not include patients under age 65, bromocriptine therapy for women for postpartum lactation suppression was not an issue.
Assembly of cohort.
The administrative databases include multiple claims for a single individual. For example, some patients with PKM were followed regularly by physicians, thereby generating multiple claims in any year; and repeat prescriptions for the same patient would each generate a separate claim. The cohort was reduced to the total number of patients without multiple counting.
A total of 138,775 physician claims with the diagnostic code for PD were identified in the 1993/94 fiscal year to create our inception cohort. We excluded 296 records (0.2%) for patients under 25. We retained records for family physicians, internal medicine specialists, and neurologists. There were only 4,147 claims (3.0%) from excluded physicians (see Methods) when patients did not have any other billings from physicians likely to treat PKM. This left 134,608 claims. There were 175,386 PKM-related drug records in the Ontario Drug Benefit (ODB) database reviewed for the 1993/94 fiscal year. Of these, 4,832 amantadine records during influenza season were excluded, leaving 170,554 claims.
These claims were generated on behalf of 22,931 patients with PKM, based on the criteria set out above. Of these, 15,306 had been similarly identified in 1992/93; those with no matches found were excluded, leaving a total of 15,306 patients for whom two or more claims in 2 years showed a primary diagnosis of PD, use of antiparkinsonian drugs, or both. A total of 7,625 patients were excluded from the cohort because they did not have claims in 1992/93.
The 15,306 patients with PKM included the following: 3,091 patients were identified with PD only on physician claims in the 2 years; 3,795 patients were identified only by having drug claims for PD medication in the 2 years; and 8,420 patients were identified with a combination of both physician claims and drug utilization. Two of these were excluded because death dates preceded service dates and coding errors were suspected, bringing the final total to 15,304.
Assembly of control subjects.
PKM cases meeting the restrictions of cohort inclusion were matched by age in years and sex with two control subjects without PKM (1:2 ratio), derived from the general population as listed in the registry of all citizens. The control subjects therefore included any living person in Ontario, matched by age and sex, who was not identified by an Ontario Health Insurance Plan (OHIP) diagnostic code for PD or who did not have an ODB claim for PD-related medication. The control subjects could have any other medical condition. Again, multiple counting did not occur since the dataset compilation utilized a unique code for each included patient.
We attempted to assess the number of cases and controls that were potentially lost from the database during the observation period. The database includes information on eligibility of persons. Of the 30,608 control subjects, 366 (1.2%) became ineligible during the study time period. Of the 15,304 cases, 139 (0.9%) became ineligible.
Mortality analysis.
The annual mortality rate was determined by the number of deaths each year adjusted to the number of surviving PKM cases and controls at the beginning of each year. Person-years of survival was calculated by weighting the number of days a person was alive during a fiscal year and expressing this as a fraction of the complete year. The sum of the total number of remaining subjects in both cohorts for each fiscal year was compiled by adding these fractions.
The analyses were performed using SAS version 6.12 (SAS Institute, Cary, NC).
Results.
The age and sex distribution of the PKM group is seen in figure 1. The table shows mortality for both PKM cases and control subjects over the 6-year period of the study. The mortality in the PKM cases (50.8%) was significantly increased compared with the control subjects (29.1%), with 1.75 being the relative risk. The survival curves for both PKM cases and controls are shown in figure 2. This shows the dramatically increased mortality for PKM cases compared with control subjects throughout the 6-year period.
Figure 1. Age/sex distribution of parkinsonian cases in inception cohort for ages over 25 years in Ontario. Gray bars = percentage male; black bars = percentage female.
Number of deaths and mortality trends in cases with parkinsonism and control subjects in Ontario by fiscal year 1993/94–1998/99
Figure 2. Survival curves of cases with parkinsonism (bottom curve) and control subjects (top curve).
As a sensitivity analysis, subgroup analyses were performed to determine if there was a bias in the estimation of the mortality OR, based on the methods used to compile the PKM and control cohorts. Patients identified in the cohort inception year by physician claims only, drug claims only, and a combination of both were analyzed separately. The mortality OR were qualitatively consistent and not different when cases were defined by physician diagnosis on payment claims only (2.43; 95% CI: 2.21, 2.68), use of PD drugs alone (2.67; 95% CI: 2.46, 2.89), or a combination of the two (2.57; 95% CI: 2.44, 2.72).
Discussion.
This population-based study suggests that PKM imposes a dramatic increase in mortality on those affected. Over a 6-year study period, 50.8% of the cases with PKM died compared with 29.1% of the control subjects. PKM cases had an overall mortality OR of 2.5 (95% CI: 2.4, 2.6) compared with the control group. The excess mortality associated with PKM did not differ appreciably over the 6 years of the study, suggesting that there was not a progressive relative increase in mortality rate in the PKM cases as their age and the duration of their condition increased.
There are several limitations to this study. No attempt was made to define the causes of death. Causes of death listed on death certificates are known to provide relatively unreliable diagnoses in patients with PD.4,8,12⇓⇓ Previous studies have identified a higher than expected mortality in PD patients due to pneumonia,8,12⇓ cerebrovascular disease,8,11,12⇓⇓ ischemic heart disease,11,12⇓ and PD per se.5,12⇓
PKM cases identified were derived from a cross-section of the population with PKM. It was not possible to identify the clinical severity, duration of disease, or specific clinical features (including dementia) from the database. However, given the population-based nature of the study, the patients included should be fully representative of the general population of patients with the disease. All patients had access to a universal health care system at no direct cost to themselves. Both the PKM patient and control populations should have received a high level of health care services. The high mortality rate observed was therefore unlikely to be due to problems with access to physicians, hospital services, or appropriate drug treatment.
We believe it is unlikely that the observed increase in mortality is due to some methodological error in this study. Only PKM patients in the cohort who had been identified as PKM cases in both 1993/94 and 1992/93 were included (see Methods)—a conservative step; nonetheless, it was one that was judged to be prudent to optimize confidence in the data. It is possible that selecting those patients who had visits in consecutive years may have biased the sample to more severe or progressive cases. Owing to misdiagnosis or miscoding by their physicians, patients may have been included who had conditions other than PD, including other forms of PKM such as progressive supranuclear palsy, multiple system atrophy, vascular-induced PKM, or drug-induced PKM. For this reason and since we do not have the potential of validating the exact clinical diagnosis in our cohort, we have presented our results relating to the diagnosis of PKM rather than PD. It has been estimated that approximately 20% of patients diagnosed with PD have other parkinsonian diagnoses on autopsy.17,18⇓ Based on the methodology, we are unable to differentiate these conditions from PD. Patients may have been included who were thought to have PD during the inception period and had this diagnosis revised at a later time. Furthermore, if patients were given antiparkinsonian treatment for reasons other than the diagnosis of PD, such as restless leg syndrome and dystonia, they might have been included inappropriately. On the other hand, our findings mirror the current population reality, including diagnostic confusion concerning PD in actual clinical practice.
The methodology we have used to assess mortality in PKM has many advantages, including the selection of a large cohort of PKM patients (15,304), exact age and sex matching of the control group at a ratio of 1:2 (30,608), with the capacity to follow the two groups over several years. The method avoids the biases inherent in some other studies, which have dealt variously with a specific clinical trial population in restricted sites,5,7,12,16⇓⇓⇓ data from specialized referral centers,4,8⇓ and small numbers of patients in community-based surveys.6,10,11,13,14,19⇓⇓⇓⇓⇓ Our results should therefore provide information that is more generalizable to the “real life” experience of the overall PKM population.
One previous publication found normalized mortality rates for patients receiving PD drug treatment.6 Prior to levodopa drug treatment, the ratio of observed to expected mortality for a small number of PD patients was 2.9 times higher, based on U.S. life tables for the same age, sex, and race.4 A differential effect on survival depending on when levodopa was introduced, suggesting early levodopa therapy increased survival of PD patients preferentially compared with later treatment, has been reported.7
Another study reported patients receiving antiparkinsonian drugs in the region of Rome.9 Their subjects were identified based on administrative records of persons receiving antiparkinsonian drugs between 1987 and 1990 and were followed on average for 5.7 years. Comparative mortality was assessed indirectly using life table comparisons, generating a standardized mortality ratio of 1.17 (95% CI: 1.14, 1.21) for PD. Another study used linked administrative databases to examine mortality associated with PD in Tayside, Scotland.15 Subjects were identified by new prescriptions for antiparkinsonian drugs in a community of 30,000 between 1989 and 1995. Control subjects from the general population were matched 9:1, and the groups were followed for up to 7 years. A smaller group from a different community was followed for up to 11 years and was included in the analysis. It is unclear how long subjects were followed, since the inception of the cohort took place throughout the study. This study found a mortality ratio of 1.76 comparing PD patients and control subjects, which is similar to our results. The current study differed methodologically from the two prior ones9,15⇓ in that we used both drug utilization profiles and physician diagnoses to define our patients and evaluated a much larger number of patients and their age- and sex-matched control subjects over a consistent 6-year period. The previous study differed in that they included only patients with a new prescription for antiparkinsonian drugs.15 This may have limited their analysis to newly treated patients only rather than the cross-sectional analysis in our study.
Increased PD mortality in a random sample of 467 elderly patients identified on door-to-door survey has been reported.19 This study observed an overall risk of death of 2.0 (95% CI: 1.6, 2.6) in the 159 subjects with PKM compared with 301 subjects without these clinical features over a period of 8 years. Another study utilized a database of general practice medical records documenting drug utilization, diagnoses, and important clinical events and found a standardized mortality ratio of 2.02 (95% CI: 1.94, 2.11) as compared with that of the general population of England and Wales.20 This study assessed only patients on levodopa, selegiline, or a combination of both drugs with the primary intent to assess differential mortality in patients taking selegiline. Our study differs in that the parkinsonian patients included all subjects with the PD diagnosis on any PD medication from the total population; the groups were also followed for 6 years as compared with 1.7 years. Three other studies have examined the effects of selegiline on the mortality of PD patients involved in clinical trials. Lees5 found a mortality ratio of 1.57 (95% CI: 1.09, 2.30) in PD patients taking selegiline in subjects requiring symptomatic therapy. The Parkinson Study Group did not find an increase in mortality in patients taking selegiline participating in the large DATATOP study.16 Another study reported an improved mortality ratio in patients taking selegiline either alone or in combination with levodopa-containing compounds.15 Importantly, all of these studies had sample sizes that were much smaller than in this report; none had the study sample drawn from a general population of this size.
The data used for these analyses were derived from three sources: prescription drug claims, physician service claims, and the registry of all citizens for public administration. These data are collected routinely by these agencies for financial or administrative purposes; they are not collected primarily for research. Williams and Young21 have summarized previous studies that examined the quality of health care administrative databases in Canada. Their overall conclusions were that demographic information is relatively complete and reliable, and although the coding systems have not kept pace with the changes in medical technology and clinical practice, billing claims for physician services are complete and relatively accurate. However, they flagged interindividual variability in diagnostic coding accuracy by physicians.22 We addressed the latter concern by excluding PD claims rendered by some types of physicians and excluding cases where the PD diagnosis rested on physician claims made in only a single year.
The increased mortality reported in this study emphasizes the large impact of the diagnosis of PKM on the health of patients with this condition. The potential benefit of neuroprotective strategies on altering the natural history of PKM is therefore large, and research into more definitive therapies for PKM should be aggressively pursued by researchers, government agencies, industry, and the voluntary sector internationally.
Appendix
Study physician inclusion and exclusive criteria.
Included study physicians: family physicians, general practitioners, neurologists, general internists, physical medicine, geriatrics, respirologists, hematologists, cardiologists, gastroenterologists, rheumatologists. Excluded study physicians: dental surgeons, pediatricians, obstetricians/gynecologists, clinical biochemists, microbiologists, thoracic surgeons, prosthodontists, plastic surgeons, oral pathologists, dermatologists, osteopaths, therapeutic radiologists, clinical immunologists, otolaryngologists, oral surgeons, diagnostic radiologists, urologists, pathologists, optometrists.
Acknowledgments
Funded jointly by an unrestricted research grant from the Parkinson’s Foundation of Canada, the National Parkinson Foundation (Miami), through a Center of Excellence Award to M.G. and the University of Toronto, and by the Institute for Clinical Evaluative Sciences.
Acknowledgment
The authors thank Anne Bain, Linda Toews, and Davida Glazer.
- Received March 27, 2001.
- Accepted September 12, 2001.
References
Letters: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Hemiplegic Migraine Associated With PRRT2 Variations A Clinical and Genetic Study
Dr. Robert Shapiro and Dr. Amynah Pradhan