The clinical course of herpes zoster: a prospective study in primary care

This paper was brought to my attention by one of its authors (Dr. Sigurdur Helgason). It is a nice illustration of an approach that is sure to become more widespread: the statistical analysis of data obtained via computerized medical records from a primary care setting.

Summary

This study from Iceland was designed to obtain an accurate assessment of the incidence and course of herpes zoster in the primary care setting, making use of the increasing availability of computerized medical record systems in ambulatory practice.

Methods

• Recruitment of subjects: Recruitment was organized around GP physicians. GP's were asked to participate if they had a computerized medical record system and agreed to the protocol. A total of 62 GP's participated, caring for about 1/3 of the population of Iceland. They were asked to report all new cases of acute Zoster in their practice, between January, 1990 and June, 1995. The diagnosis of Zoster was on clinical criteria alone; patients gave verbal consent to participate and were excluded if there was cognitive impairment precluding an accurate history.
   
• Data collected: Upon notification of a case, one of the authors of the study contacted the patient and ascertained data about the rash, associated symptoms, treatment and other medical conditions. Patients were followed-up by telephone at least 1, 3 and 12 months after initial appearance of the rash and more frequently if there was postherpetic neuralgia.

• Patients recruited: 457 patients were included in the study, after exclusion of 48 patients (38 for incorrect diagnosis as judged by the authors). Most patients were reported by the GP's participating in the study (75%), but 25% were recruited by reviewing computer printouts of medical records. The number of person-years observed was 229,547.
   
• Incidence of zoster: Overall 2.0 per 1000 person years. Age-standardized in the over 60 population: males 3.6 and females 5.6 per 1000 years. Overall distribution (not age-standardized) bimodal, with one quarter of the cases in the population under 20 years of age and 34% in the 60 and over population.
   
• Dermatomes affected: Thoracic (62%), lumbar (14%), cervical (11%), ophthalmic (8%), others (5%).
   
• Recurrent zoster: One percent during the study period, 5% including prior history.
   
• Associated conditions: Diabetes in 13 patients (2%), malignancies within 6 months of zoster in 5 patients (1%), HIV in one and steroid or cyclosporine therapy in 5 (1%).
   
• Postherpetic neuralgia:
   
• At one month after onset (information was available for 370 of 462 episodes):
• Age under 40: 98% were pain free, 2% had mild pain.
• Age 40-60: 79% were pain free, 21% had mild/moderate pain.
• Age over 60: 59% were pain free, 37% had mild/moderate pain, 3% had severe pain
   
• At three months after onset (information available for 427 of 462 episodes):
• Age under 40: 99.5% were pain free, 0.5% had mild pain.
• Age 40-60: 95% were pain free, 5% had mild pain.
• Age over 60: 81% were pain free, 18% had mild/moderate pain, 1% had severe pain
   
• At twelve months after onset (information available for 457 of 462 episodes):
• Age under 40: 100% were pain free.
• Age 40-60: 98% were pain free, 2% had mild pain.
• Age over 60: 92% were pain free, 8% had mild/moderate pain.
   
• Severity / complications / treatment: Ophthalmic zoster developed in 39 patients, 14 had ocular involvement. Eighteen patients were hospitalized for zoster (4%). Only 17 patients (4%) were treated with systemic acyclovir, mainly because of age and severity of rash.

The authors compare their results to those of other studies, which have shown largely comparable findings. They note that their incidence of zoster in the younger population is higher than previously noted and that their incidence of post-herpetic neuralgia is similar to earlier studies but lower than that found in the placebo arm of recent trials of acyclovir. They estimate that a GP caring for 2000 patients can expect to see zoster four times a year and post-herpetic neuralgia lasting more than three months every other year, principally in the older population. Finally, in their series zoster was rarely associated with major illnesses.

Comment

Beyond the data presented here, the concept of using outpatient computerized medical records to obtain data for such a study is an important one and raises a number of interesting issues:

• Electronic medical record systems offer the potential to obtain reams of incidence data with relative ease. For this data to be accurate, however, the population being observed or the population at risk needs to be clearly defined. Since, in many countries, patients shift between health care providers, it is necessary to have a unique patient identifier that is used when analyzing such data. As noted earlier, if a large number of patients are potential clients of a physician but not in that physician's EMR system, the population at risk will be under-estimated and incidences over-estimated.
   
• If computerized medical record systems contain data about patients that is coded differently or is in other ways not adequately standardized, it will be difficult to pool information. When pooling data from different record systems, we will, in general, be limited to making statements about data that is present in all of them. We are, in a sense, limited to the lowest common denominator when pooling such data. Some sort of standardization, a minimal data-set, is urgently needed.
   
• As a corollary to the above, a "Rosetta Stone" or meta-medical record which would allow mapping of the main components of every type of EMR to a standard record would be very useful and would permit cross-platform analysis of data from multiple systems. Each vendor of an EMR system would then be responsible for providing a mapping of its data to the standardized meta-medical record. This would also allow easy portability of a patient's computerized chart from one system to another.
   
• Enabling private practitioners to provide data for studies such as this one has significant benefits. It allows health care researchers to conduct studies in the "real world" with greater ease, while it also allows practitioners in the trenches of primary care to participate in research in a meaningful way. This could be an invigorating and gratifying experience for both sides.
   
• On a more cautionary note, the number of studies that can and will be performed using this approach is practically limitless. Once a mechanism is in place for accessing and exploiting the data present in private practitioners' offices, that data can be "massaged" ad infinitum. Those with a particular axe to grind or potion to sell will be able to sit at their computers and come up with statistics to back them up. If the calcium channel blocker industry gets annoyed at bad publicity, it can run "ACE inhibitor prescriptions" against every disease in the ICD-9 book until a correlation pops up that can then be "investigated" and publicized. Just as post-hoc subgroup analysis is subject to skepticism, we will, in the long run, need criteria for the evaluation of statistically derived outpatient data.

Reader comments

First: I want to emphasize that the main reason and actually the only reason for this study was to find the point prevalence (traditionally called incidence in zoster studies of this kind) of postherpetic neuralgia at different time points after acute zoster in a primary care population. We have an age breakdown of 90% of the persons under observation, or a third of Icelanders. That is, nearly all practices could give the age distribution of the patients they cared for (registered). After comparing this with the age distribution of the total population (just over 261 thousand in 1992) we found it to be almost the same. We extrapolated from the 90% available, assuming the practices providing age strata were similar to those not providing this information.

Secondly: The number of person years was determined by multiplying the number of patients registered with each practice or GP by the duration of time each particular practice or GP participated in the study. Bias in estimating incidence due to a possible difference between catchment area/population and the population actually registered ("belonging") to a particular GP is not likely. The reason is that to be included as a case of zoster the patient had to be registered with a participating GP. All patients diagnosed with zoster but whose GP was not in the study were excluded, even if a GP in the study diagnosed them (during call or other services). This ment that GP´s not in the study, but who knew of it, often contacted me telling me about new cases in their care which I had to refuse, of course. One more point is that we kept a low profile with this study because we wanted only patients who normally would have sought medical attention for their zoster (avoiding publicity/referral bias).

Thirdly: I agree that 6 months is hardly enough follow up with malignancy in mind. What we did was to follow all for 12 months. I am a little obsessive when it comes to follow-up and those moving abroad were almost all eventually reached via phone. 99% follow-up over a year must be a record. The time frame of 6 months prior to or after the zoster episode was a collective decision. Actually, looking back at the data and extending this to 12 months after the zoster would not have changed the actual number of patients with malignancy "associated" with zoster.

Finally: Of the 14 patients still having pain after 12 months I have contacted eleven, 24 to 52 months after the zoster episode. One of the two patients with moderate pain was unchanged, but the other one defined his pain as mild and improving. Of the 12 patients with mild pain, two were worse, three better (pain free), four unchanged and three have not been reached/evaluated.

Sincerely,
Sigurður Helgason

That pretty much clarifies all the questions I had. Thanks for taking the time to answer in so much detail. -- mj

Date: Sun, 04 Jan 1998
From: Intrade <intrade@worldnet.att.net>
Organization: Intrade

Sir,

In your study did you find any incidence of Herpes Zoster during pregnancy?
If so, how was it treated?
Did Herpes Zoster and/or its treatment lead to any birth defects?

Sincerely,

Manish H. Shah.
 

Dr. Sigurdur Helgason, study author, responds:

Date: Thu, 8 Jan 1998
From: Sigurdur Helgason <sh@centrum.is>

My short answer without going to the original data is:

Six women were known to be pregnant. None were treated and in no case were there any complications - maternal or foetal. At follow up (patient contact 12 months after the zoster and GP contact about 24 months after the zoster) there was no indication of complications. Although we should keep in mind that we did not examine babies or ask focused questions in that direction as it is our experience that zoster in pregnancy is a benign condition.

Sincerely yours Sigurður Helgason.

I read with interest your perspective on clinical data bases in physician offices.  Conceptually this is where we all should be heading; however the US is still struggling with continuity of disease classifications at the hospital level.  So much of the classification process is geared toward reimbursement issues instead of clinical pertinence.  While it seems insurmountable to standardize coding, it's what is needed to control costs and improve patient care.

Sincerely
D. Jan Powell, MBA, RRA, CCS &
BS - Telecommunications Management
Health Information Services Consultant
 

November 12, 2000

Long term follow-up results of this study were published in the September 30, 2000 issue of the British Medical Journal, and are consistent with the results published in this paper.

Journal Club on the Web main screen

Submit a comment about this article
 
 

Note