## Abstract

Acute pyelonephritis causes significant morbidity, tends to recur, and can be fatal; however, little is known regarding its epidemiology. In this paper, the authors describe the epidemiology of acute pyelonephritis in South Korea by using nationwide heath insurance claims data from 1997 to 1999. The National Health Insurance System of South Korea covers almost the entire population (99%). The overall average annual incidence rate of pyelonephritis in 1997–1999 was 35.7 per 10,000 population (male, 12.6; female, 59.0). Approximately one of every seven patients was hospitalized (incidence per 10,000: inpatients, 5.5; outpatients, 30.1). Incidence varied with age and was higher in the summer season. Following an initial episode, the risk of a second episode within 12 months was 9.2% for females and 5.7% for males; by contrast, the risk of a fifth episode within a year following a fourth episode was 50.0% for females and 53.0% for males. Female sex (hazard ratio = 1.89, 95% confidence interval: 1.60, 2.23), advancing age, outpatient treatments (hazard ratio = 1.35, 95% confidence interval: 1.14, 1.60), and medical aid (hazard ratio = 1.23, 95% confidence interval: 1.08, 1.40) increased the risk of any recurrence. Pyelonephritis has a clear seasonal pattern and high rate of recurrence. The incidence of hospitalization for pyelonephritis in South Korea is similar to that in the United States and Canada.

Received for publication April 1, 2004; accepted for publication June 3, 2004.

Acute pyelonephritis is a kidney parenchyma infection that occurs most often as a result of ascending infection from the bladder (1). Clinical manifestations include fever, flank pain, and costovertebral angle tenderness that may be accompanied by symptoms of cystitis (frequent and painful urination). A 1997 survey utilizing the US 1997 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS) estimated the incidence of hospitalized pyelonephritis to be 11.7 cases per 10,000 population among females and 2.4 cases per 10,000 population among males (2). Although pyelonephritis responds well to antibiotics, increasing antimicrobial resistance and concomitant reduced clinical efficacy have been observed (3). Infants, the elderly, and pregnant women with acute pyelonephritis are most often hospitalized (4), and in-hospital mortality rates are less than 2 percent (females, 1.65 percent; males, 0.73 percent) (2). However, almost one quarter of females hospitalized with acute pyelonephritis have a second episode within 6 months (5), and multiple recurrences have been reported. Despite the relatively high frequency and severity of this infection, the basic epidemiology of acute pyelonephritis has not been described to our knowledge. There are several review papers about pyelonephritis but few original publications estimating hospitalization or in-hospital mortality rates (2, 6, 7). We found no incidence or recurrence estimates that included outpatients.

In this paper, we describe the incidence, recurrence, seasonality, mortality, and medical costs of acute pyelonephritis for inpatients and outpatients. We used 1997–1999 South Korean nationwide health insurance claims data.

## MATERIALS AND METHODS

### Data source

#### National health insurance data

The National Health Insurance System of South Korea covers almost the entire population, including management of medical aid for low-income families (98.9 percent coverage in 1998) (8). We obtained insurance claims data from the National Health Insurance Corporation, with patient name and the last two digits of the national registration number removed to ensure patient confidentiality. Patient information included the modified national registration number, date of birth, gender, insurance corporation code, insurer type, clinic or hospital code, area of the clinic or hospital, care type (inpatient/outpatient), treatment start date, number of treatment days, admission or visit day, primary discharge diagnosis code, subdischarge diagnosis code, treatment department, and medical costs covered by health insurance. Medical costs did not include copayment by patients, food expenses, single-room charge, magnetic resonance imaging, or other new techniques. The discharge diagnoses were coded by using the International Classification of Diseases, Tenth Revision (ICD-10).

#### Mortality data

National mortality data in South Korea have been computerized since 1983. Mortality data originate from hospital medical records departments and the statistics division of the provincial government offices. Management of provincial statistics is ongoing, with central management in the Vital Statistics Division of the Korean National Statistical Office. Before national mortality data are provided for each year, the patient name and the last digit of the national registration number for each death record are removed to ensure patient confidentiality. Death records contain patient name, address, date of birth, national registration number, gender, date of death, cause of death, occupation and educational level, location at time of death, marital status, diagnosing provider, date that the death was reported, and place of death. Cause of death is coded by using ICD-10. In addition to the primary cause of death, a secondary cause of death is recorded for a limited number of records.

### Definitions

We defined a case as any person whose primary discharge diagnosis was acute pyelonephritis: ICD-10 codes N10 (acute tubulo-interstitial nephritis) and N12 (tubulo-interstitial nephritis, not specified as acute or chronic). Episodes without a pyelonephritis claim during the previous year were considered sporadic; episodes with a pyelonephritis claim during the previous year were considered recurrent. When the time period between two episodes was more than 365 days, the second episode was considered sporadic. Two or more claims during a 14-day period were considered a single episode. Details of case selection criteria are shown in figure 1.

As noted, the first and second episodes for case 1 were sporadic because no episodes occurred during the previous year. The third episode for case 1 was a first recurrence. The last episodes for cases 2 and 5 were excluded from the recurrence rate analysis because they occurred after September 1999, which was defined as the end of the observation period. The first episodes for cases 4–7 could not be classified as sporadic or recurrent. Recurrent episodes after undefined episodes were also not defined and were excluded from the recurrence analyses.

The observation period used to calculate recurrence rates was defined as January 1, 1997, to September 30, 1999. We used this period because claims data for any given year are incomplete for treatment claims occurring in November and December of that same year because there is a 1- to 2-month lag in billing.

### Data analysis

Incidence rates were calculated by using data from the 1997–1999 South Korea Census Bureau population projection for the denominator (46,286,503 in 1998) (9). Incidence rates by health insurance versus medical aid were calculated by using the covered population as the denominator (44,472,000 for health insurance and 1,322,696 for medical aid in 1998) (10). To verify seasonal variation, we used the Poisson dispersion test (11).

Using the national registration number, we linked the insurance data and national mortality data to calculate the mortality rate. We used the number of deaths within 1 month following treatment of pyelonephritis as the numerator (from national mortality data) and the number of pyelonephritis cases in 1997–1998 for the denominator (from insurance claims data).

To calculate the hazard ratios for recurrences, we used a Cox proportional hazards regression analysis that did not adjust for repeated recurrent events within a person. To adjust for repeated recurrent events within a person, we used a frailty model (12). We included all variables and tested all possible interactions; only significant interactions were included in the final model. For the analysis of multiple recurrences, the variable “previous interval of recurrence” was added to the model. Hazard ratios were calculated by exponentiating the coefficient.

Claims-based insurance data were linked to national mortality data by using national registration numbers. The Korean national registration number has 13 digits: six for birth date, one for gender, four for birthplace code, and one for birth order for babies who are born in the same place, are born on the same date, and are of the same gender. The last digit is a function of the previous 12. Therefore, we lacked information on only one informative digit, that indicating birth order. Since the number of possible birthplaces is 3,519 and the number of births per year is about 700,000, on average 0.27 births occur in each birthplace per day (700,000/(365 days × 2 genders × 3,519 birthplaces)). Therefore, when the last two digits of the national registration number are excluded, the probability of error in matching is very small: an estimated 0.93 × 10–6 (–0.27 × 0.00357 × 0.27 × 0.00357, where the annual incidence rate of pyelonephritis is 0.00357 according to our results).

All analyses were conducted by using the SAS (version 8.1) software program (SAS Institute, Inc., Cary, North Carolina) and R program (13). Analysis of these data were approved by the Institutional Review Board of the University of Michigan on December 20, 2002.

## RESULTS

### Incidence

There were 164,661 claims for acute pyelonephritis in 1997, 166,436 in 1998, and 165,192 in 1999 (total = 496,289). Most (89 percent) were for acute tubulo-interstitial nephritis.

The annual incidence of acute pyelonephritis per 10,000 South Korea population was 35.7. The incidence of hospitalization for acute pyelonephritis was 9.96 per 10,000 for females and 1.18 per 10,000 for males. Outpatient incidence was higher: 49.1 per 10,000 for females and 11.4 per 10,000 for males. Approximately one of every seven patients was hospitalized (inpatients, 5.5 per 10,000; outpatients, 30.1 per 10,000). For every age after age 2 years, females had a higher incidence than males did, with average overall incidence rates of 59.0 per 10,000 for females and 12.6 per 10,000 for males. Sporadic episodes among females increased dramatically after 15 years of age until age 25 years. After age 25 years, incidence remained stable until 80 years of age (figure 2). By contrast, the incidence among males increased gradually with age.

The ratio of outpatient to inpatient treatment was higher for males than for females but decreased with age for both genders (table 1). Both inpatient and outpatient incidence varied by payment source. Females and males who received medical aid (government aid for lower income families) constituted more inpatients and fewer outpatients than overall. We found no differences in incidence by hospital location. Incidence also varied by hospital size, with larger hospitals having a higher incidence of inpatients. Internal medicine (70.9 percent for females and 55.4 percent for males) or urology (9.5 percent for females and 17.3 percent for males) specialists cared for most persons treated as inpatients.

The hospitalization rate per 100 claims was higher for females of every age, with the exception of those aged 5 years or younger. The hospitalization rate increased dramatically at ages 15–29 years for both genders, followed by a more gradual rise with age. The difference between females and males also increased slightly with age (figure 3).

### Seasonality

Pyelonephritis incidence peaked in the summer season (July or August) for both females and males and in all study years (Poisson dispersion test, p < 0.001) (figure 4). There were geographic differences in incidence, ranging from 26 to 45 per 10,000, but we observed summer peaks in each area. Southern areas showed a higher incidence and earlier peak times, but there was no clear pattern by latitude, and differences were not statistically significant.

### Recurrence

To separate recurrent from sporadic episodes, we used data from the previous year (refer to the Materials and Methods section). Thus, this analysis is limited to selected cases of pyelonephritis occurring in 1998 and 1999 (182,315 sporadic and 20,916 recurrent episodes). The pattern of incidence of recurrent episodes was similar to that for sporadic episodes, but the average rate was 7.0 per 10,000, one fifth of the sporadic-episode incidence. For males, the incidence of both sporadic and recurrent episodes increased continuously with age, with average incidences of 9.0 and 1.0 per 10,000, respectively.

The probability of recurrence increased with the number of previous recurrences. Following a sporadic episode, the risk of a recurrence within 12 months was 9.2 percent for females and 5.7 percent for males. Among persons with four previous episodes, the risk of a fifth episode within 12 months was 53.0 percent for females and 50.0 percent for males (table 2). The recurrence rate varied by interval of previous recurrence; the chance of a third recurrence for those with a second recurrence within 61 days was 41.1 percent; however, if the time period between the first and second recurrences was 121 days, the risk of a third recurrence was 24.1 percent. Median intervals between recurrences shortened with frequency of recurrence for both males and females. One male (aged 42 years) experienced 27 episodes in 3 years. Among sporadic cases, one woman aged 45 years experienced 14 recurrences within 17 months, which was the maximum number of recurrences recorded.

To identify risk factors for recurrence, we used a Cox proportional hazards ratio analysis, and a frailty model, which can adjust for multiple recurrences. Both of the models considered all variables and all possible interaction terms simultaneously. The results of the Cox analysis were similar to those of the frailty model, but the hazard ratios were slightly higher. Unadjusted results are not shown in this paper.

The hazard ratio for any recurrence increased with age and was significantly higher for females, outpatients, and those receiving medical aid (table 3). There was a significant interaction between gender and care type: when male inpatients were considered the reference group, male outpatients, female inpatients, and female outpatients had 1.35, 1.89, and 1.99 times higher hazards for recurrence, respectively. Hospital size also made a difference: when those treated as inpatients in hospitals (≥30 beds) were used as the reference group, the hazard ratios for hospital outpatients, clinic (<30 beds) inpatients, and clinic outpatients were 1.35, 1.12, and 1.99, respectively. Seasonal differences in the hazard ratio were smaller, with slightly higher hazards for the winter and spring seasons than for the fall season.

For those experiencing two or more recurrences, it was possible to predict the risk of subsequent recurrences by the interval between previous recurrences. Those with shorter previous intervals had higher hazard ratios for a subsequent recurrence. In this group, the hazard ratios were not increased by gender and payer type; however, for those in the age group 65 years or older, there was a higher hazard ratio of recurrence, which also varied by hospital size. When hospital (≥30 beds) inpatients were considered the reference group, the hazard ratios for clinic (<30 beds) inpatients, hospital outpatients, and clinic outpatients were 1.66, 1.64, and 1.78, respectively (table 3).

### Mortality

The overall mortality rate per 1,000 pyelonephritis cases was 1.2 (females, 1.0; males, 2.0). When we considered just renal or infectious disease as the cause of death noted on the death certificate, the mortality rate was 0.2 per 1,000 for females and 0.3 per 1,000 for males. Among hospitalized patients, the average mortality rate was 2.1 per 1,000 (females, 1.7; males, 5.3). For renal or infectious disease as the cause of death, the mortality rate for hospitalized cases was 0.5 per 1,000 for females and 0.6 per 1,000 for males. Renal or infectious disease accounted for 18 percent, cancer for 21 percent, cardiovascular disease for 19 percent, senility for 10 percent, diabetes complications for 8 percent, and gastrointestinal disease for 7 percent of all deaths occurring in 1 month following acute pyelonephritis.

### Treatment duration and costs

Treatment duration per episode (8.4 days) was similar for males and females, averaging 7.2 days for outpatients and 14.1 days for inpatients. The average number of visits per outpatient episode was 2.9. The average duration of hospitalization per inpatient episode was 7.9 days. The 95 percent confidence intervals for treatment duration were very close to the mean.

The medical cost (clinic charge per physician visit + cost per prescription + hospitalization charge) increased each year. Therefore, the average cost was calculated by using the most recent year, 1999. The average medical cost per episode for hospitalized patients was $513.60 (females,$511.00; males, $535.40). Among outpatients, the average medical cost per episode was$57.30 (females, $59.00; males,$50.70). The total medical costs for acute pyelonephritis were $15.1 million,$16.9 million, and $17.6 million in 1997, 1998, and 1999, respectively. The cost for pyelonephritis was 0.2 percent of the national total medical cost for 1999 ($9.75 billion).

## DISCUSSION

This is the first report we are aware of describing the epidemiology of acute pyelonephritis that includes information on incidence among inpatients and outpatients of both genders. The epidemiology of acute pyelonephritis is similar to that of urinary tract infection overall: higher incidence in females than males, a sharp increase at puberty, and a high rate of recurrence (14). For this most severe form of urinary tract infection, we also observed a clear seasonal trend in incidence among both females and males.

Age and gender differences in our data were similar to those in previous studies (2, 6). The hospitalization rate per 10,000 population among females (9.96) and males (1.18) was slightly lower than that reported for the United States (females, 11.7; males, 2.4) but higher than that for US Asian/Pacific Islanders (females, 5.4; males, 0.9) (2). There were no comparable data regarding outpatient incidence rates. The outpatient incidence was five times higher than the inpatient incidence for females and 10 times higher for males.

Recurrence of pyelonephritis is a significant problem, accounting for 15.2 percent of all acute pyelonephritis episodes occurring in South Korea in 1998. The frequency of recurrence was consistent with that in the one previous report we found (5). This study, conducted in 1973 among females aged 15–35 years hospitalized for pyelonephritis, found a 6-month recurrence of 23 percent (16 per 74) and a 3-year recurrence of 40 percent (29 per 74) (5).

Female sex was the most important risk factor for recurrence. This finding might be partially explained by anatomic differences, which are known to cause a higher incidence of all urinary tract infections in females. The observed variation in recurrence rates by medical care type, payer type, and level of hospital (number of beds) may relate to treatment type, duration, or both, which may modify the risk of recurrence. Alternatively, underlying morbidities and disease severity may also play a role. Some underlying conditions, such as diabetes, may predispose to pyelonephritis (6).

Once a person experienced multiple recurrences, only the previous interval between pyelonephritis episodes, an age of 65 years or older, and the level of hospital significantly predicted a subsequent recurrence; gender had no effect. This finding suggests that the time between recurrences may be a potentially useful clinical and research marker. Some persons have a propensity to recurrence, and some bacterial strains may be harder to clear either in general or in particular for a host with a propensity to recurrence (15). While it is impossible to determine with these types of data, it might be that different types or durations of treatments may be indicated for those with a short time between recurrences. A recent study suggested a positive association of history of urinary tract infections with development of renal cell carcinoma, so identifying treatments to minimize recurrences may decrease not only associated morbidity but also potential sequelae (16).

We observed a strong seasonal trend in acute pyelonephritis incidence. Seasonal trends provide useful clues to understanding transmission patterns. Explanations for seasonality can be grouped into three types: pathogen appearance and disappearance, environmental changes, and host-behavior changes (17). Seasonal variation in acute pyelonephritis may be due to increased host resistance in the winter season or decreased virulence of the pathogen during that time. Escherichia coli, the most common cause of pyelonephritis and all urinary tract infections, are transmitted by person-to-person direct contact and by contaminated food or water. Presumably, uropathogenic E. coli are also transmitted by these modes, but the relative contributions of each mode are uncertain (15). Transmission by contaminated water might contribute to a summer peak of pyelonephritis. However, the only known documented seasonal variation in urinary tract infection caused by Staphylococcus saprophyticus has a late autumn peak (18).

In 1997, in-hospital mortality in the United States was 7.3 per 1,000 hospitalizations for females and 16.5 for males (2). These mortality rates were two to three times higher than our results (females, 1.7; males, 5.3). However, in both sets of data, the mortality rate for males was higher. Average treatment duration for outpatients (7.2 days) and for inpatients (14.1 days) was similar to the treatment duration recommended by the Infectious Disease Society of America (19). However, since we did not have data about antibiotics and complications, it is difficult to comment on the appropriateness of the treatment duration.

A strength of our study was the use of national health insurance data, a very large and comprehensive database that enabled us to analyze trends for both inpatients and outpatients. Estimates of mortality required linking mortality with insurance claims data, which may lead to errors in linkage, but we estimate that these errors were very small (refer to the Materials and Methods section). However, a problem with using administrative data to measure population-based rates of disease is the reliability of the diagnoses. Nicolle et al. (6) reported that 80 percent of patients with a discharge diagnosis of acute pyelonephritis met audit criteria for admission for acute pyelonephritis. In another study using South Korea health insurance data, the validity of diagnosis for aseptic meningitis in children was 88 percent (20). Since clinical symptoms, along with urinalysis or bacterial cultures, were not used to define the diagnosis, it is possible that the results were biased toward an overestimation of the incidence of acute pyelonephritis because other types of acute tubulo-interstitial nephritis cannot be excluded (e.g., drug induced). If information on treatment type (e.g., antibiotics) were available, this bias might be improved.

By contrast, additional cases may have been coded under other diagnoses, such as urinary tract infection or septicemia, which would lead to an underestimation of the true rates. Furthermore, until July 2000, patients in South Korea could obtain antibiotics at a drugstore without prescription, which might lead to an underestimation of the diagnosis of acute pyelonephritis, especially of recurrence. Following the institution of prescriptions for antibiotics, the number of claims for acute tubulo-interstitial nephritis (ICD-10 code N10) increased about 10.4 percent in 2001 and 13.6 percent in 2002 (21, 22)compared with the average number of claims during 1997–1999.

Nonetheless, we think that, overall, the rates of acute pyelonephritis based on the health insurance data are likely to be a reasonable approximation of the true rates. We used ICD-10 codes N10 (acute tubulo-interstitial nephritis) and N12 (tubulo-interstitial nephritis, not specified as acute or chronic) to define acute pyelonephritis because multiple recurrent cases can be coded as N12 and this selection of codes is consistent with other research (2). In conclusion, acute pyelonephritis occurs frequently in females and among males and females aged 60 years or older. The incidence among females treated as outpatients was five times higher than that for females treated as inpatients, and 10 times higher for males. Recurrences were common and occurred more frequently in females, those aged 45 years or older, those treated as outpatients, and persons receiving medical aid rather than having health insurance. The recurrence rate increased both with frequency of recurrence and shorter intervals between recurrences. There was a clear summer seasonal trend. Further research regarding the frequent recurrent group and the seasonality pattern is needed.

## ACKNOWLEDGMENTS

The first author’s work was supported by the BumSuk Academic Research Fund of 2003. The second author’s work was supported in part by the Korea Science and Engineering Foundation (KOSEF) through the Statistical Research Center for Complex Systems at Seoul National University. The fourth author’s effort was supported in part by R01 DK55496.

Correspondence to Dr. Moran Ki, Department of Preventive Medicine, Eulji University School of Medicine, 143-5, Yongdu-Dong, Joong-Gu, Daejeon, 301-832, South Korea (e-mail: kimoran@eulji.ac.kr).

FIGURE 1. Examples of criteria for selecting pyelonephritis cases, by definitions, South Korea, 1997–1999. Sporadic episodes: all episodes without an instance of pyelonephritis reported during the previous year; recurrent episodes: all episodes with an instance of pyelonephritis reported during the previous year. For a more detailed discussion of cases 1–7, refer to the Materials and Methods, Definitions section of the text.

FIGURE 1. Examples of criteria for selecting pyelonephritis cases, by definitions, South Korea, 1997–1999. Sporadic episodes: all episodes without an instance of pyelonephritis reported during the previous year; recurrent episodes: all episodes with an instance of pyelonephritis reported during the previous year. For a more detailed discussion of cases 1–7, refer to the Materials and Methods, Definitions section of the text.

FIGURE 2. Incidence of sporadic vs. recurrent episodes of acute pyelonephritis, by age, among females and males in South Korea, 1998. Sporadic episodes: all episodes without an instance of pyelonephritis reported during the previous year; recurrent episodes: all episodes with an instance of pyelonephritis reported during the previous year.

FIGURE 2. Incidence of sporadic vs. recurrent episodes of acute pyelonephritis, by age, among females and males in South Korea, 1998. Sporadic episodes: all episodes without an instance of pyelonephritis reported during the previous year; recurrent episodes: all episodes with an instance of pyelonephritis reported during the previous year.

FIGURE 3. Hospitalization rate of pyelonephritis per 100 claims, by gender and age, South Korea, 1997–1999.

FIGURE 3. Hospitalization rate of pyelonephritis per 100 claims, by gender and age, South Korea, 1997–1999.

FIGURE 4. Incidence of acute pyelonephritis, by treatment month, South Korea, 1997–1999.

FIGURE 4. Incidence of acute pyelonephritis, by treatment month, South Korea, 1997–1999.

TABLE 1.

Average annual incidence rate of acute pyelonephritis, by gender and location of treatment, per 10,000 population, South Korea, 1997–1999

 Females Males Inpatient Outpatient Inpatient Outpatient Age (years) 0–19 2.38 14.19 0.72 7.20 20–39 13.46 64.42 0.91 11.05 40–59 11.52 67.44 1.51 14.61 60–79 15.37 57.15 3.09 20.51 ≥80 11.20 28.07 4.93 25.10 Payer type Health insurance 9.75 49.63 1.16 11.68 Medical aid* 13.05 19.40 2.25 8.19 Hospital location City 10.07 49.74 1.13 11.50 Other 9.99 49.14 1.25 11.66 Hospital level (no. of beds) General hospital (≥100) 6.45 10.00 0.79 3.43 Small-sized hospital (30–99) 2.39 4.82 0.26 1.39 Clinic (<30) 1.08 34.13 0.13 6.57 Public health center 0.03 0.14 0.00 0.06 Total 9.96 49.09 1.18 11.45
 Females Males Inpatient Outpatient Inpatient Outpatient Age (years) 0–19 2.38 14.19 0.72 7.20 20–39 13.46 64.42 0.91 11.05 40–59 11.52 67.44 1.51 14.61 60–79 15.37 57.15 3.09 20.51 ≥80 11.20 28.07 4.93 25.10 Payer type Health insurance 9.75 49.63 1.16 11.68 Medical aid* 13.05 19.40 2.25 8.19 Hospital location City 10.07 49.74 1.13 11.50 Other 9.99 49.14 1.25 11.66 Hospital level (no. of beds) General hospital (≥100) 6.45 10.00 0.79 3.43 Small-sized hospital (30–99) 2.39 4.82 0.26 1.39 Clinic (<30) 1.08 34.13 0.13 6.57 Public health center 0.03 0.14 0.00 0.06 Total 9.96 49.09 1.18 11.45

* A national government health-care system for low-income families.

TABLE 2.

Recurrence rate of pyelonephritis within 1 year following an initial sporadic episode, by frequency of recurrence, previous interval of recurrence, and gender, South Korea, 1998–1999

 Order of episodes No. of episodes Median interval of recurrence (days) Recurrence rate (%) Recurrence rate (%) by previous interval of recurrence <61 days 61–120 days ≥121days Females Sporadic 145,667 1st recurrence 13,419 59 9.2 2nd recurrence 2,916 56 21.7 24.2 22.1 17.8 3rd recurrence 1,008 46 34.6 41.1 34.4 24.1 4th recurrence 417 39 41.4 49.4 34.1 26.8 5th recurrence 221 33 53.0 58.7 45.7 41.8 Males Sporadic 36,648 1st recurrence 2,104 44 5.7 2nd recurrence 455 42 21.6 23.3 18.1 19.8 3rd recurrence 155 37 34.1 36.3 33.7 28.4 4th recurrence 78 32 50.3 61.2 37.5 27.3 5th recurrence 39 35 50.0 55.6 38.5 60.0
 Order of episodes No. of episodes Median interval of recurrence (days) Recurrence rate (%) Recurrence rate (%) by previous interval of recurrence <61 days 61–120 days ≥121days Females Sporadic 145,667 1st recurrence 13,419 59 9.2 2nd recurrence 2,916 56 21.7 24.2 22.1 17.8 3rd recurrence 1,008 46 34.6 41.1 34.4 24.1 4th recurrence 417 39 41.4 49.4 34.1 26.8 5th recurrence 221 33 53.0 58.7 45.7 41.8 Males Sporadic 36,648 1st recurrence 2,104 44 5.7 2nd recurrence 455 42 21.6 23.3 18.1 19.8 3rd recurrence 155 37 34.1 36.3 33.7 28.4 4th recurrence 78 32 50.3 61.2 37.5 27.3 5th recurrence 39 35 50.0 55.6 38.5 60.0
TABLE 3.

Hazard ratios of recurrent episodes of acute pyelonephritis during a 12-month period,* South Korea, 1998–1999

 Any recurrences† Multiple recurrences‡ Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Gender Male 1.0 1.0 Female 1.89 1.60, 2.23 0.98 0.57, 1.70 Age (years) 0–14 1.0 1.0 15–44 1.11 1.03, 1.19 0.98 0.84, 1.15 45–64 1.27 1.17, 1.37 1.15 0.97, 1.35 ≥65 1.28 1.17, 1.39 1.32 1.10, 1.58 Care type Inpatient 1.0 1.0 Outpatient 1.35 1.14, 1.60 1.64 0.96, 2.81 Season Fall 1.0 1.0 Winter 1.10 1.05, 1.15 0.97 0.88, 1.06 Spring 1.10 1.05, 1.15 1.01 0.93, 1.10 Summer 1.02 0.97, 1.06 0.97 0.89, 1.06 Area of hospital City 1.0 1.0 Other 0.98 0.95, 1.01 0.93 0.87, 0.99 Level of hospital Hospital 1.0 1.0 Clinic 1.12 0.97, 1.25 1.66 1.11, 2.46 Payer type Health insurance 1.0 1.0 Medical aid§ 1.23 1.08, 1.40 0.94 0.70, 1.28 Interaction Gender × care type 0.78 0.66, 0.93 1.03 0.59, 1.80 Hospital level × care type 1.32 1.14, 1.52 0.66 0.44, 0.98 Previous interval of recurrence¶ (days) <61 1.0 61–120 0.85 0.78, 0.92 ≥121 0.79 0.73, 0.85
 Any recurrences† Multiple recurrences‡ Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Gender Male 1.0 1.0 Female 1.89 1.60, 2.23 0.98 0.57, 1.70 Age (years) 0–14 1.0 1.0 15–44 1.11 1.03, 1.19 0.98 0.84, 1.15 45–64 1.27 1.17, 1.37 1.15 0.97, 1.35 ≥65 1.28 1.17, 1.39 1.32 1.10, 1.58 Care type Inpatient 1.0 1.0 Outpatient 1.35 1.14, 1.60 1.64 0.96, 2.81 Season Fall 1.0 1.0 Winter 1.10 1.05, 1.15 0.97 0.88, 1.06 Spring 1.10 1.05, 1.15 1.01 0.93, 1.10 Summer 1.02 0.97, 1.06 0.97 0.89, 1.06 Area of hospital City 1.0 1.0 Other 0.98 0.95, 1.01 0.93 0.87, 0.99 Level of hospital Hospital 1.0 1.0 Clinic 1.12 0.97, 1.25 1.66 1.11, 2.46 Payer type Health insurance 1.0 1.0 Medical aid§ 1.23 1.08, 1.40 0.94 0.70, 1.28 Interaction Gender × care type 0.78 0.66, 0.93 1.03 0.59, 1.80 Hospital level × care type 1.32 1.14, 1.52 0.66 0.44, 0.98 Previous interval of recurrence¶ (days) <61 1.0 61–120 0.85 0.78, 0.92 ≥121 0.79 0.73, 0.85

* Multivariate frailty model adjusting for repeated recurrent events. All variables listed were included in the model.

† This analysis included all recurrences following sporadic episodes. A sporadic episode had no instance of pyelonephritis reported during the previous year (refer to the text for further information).

‡ This analysis was limited to persons experiencing two or more recurrences of pyelonephritis over 21 months.

§ A national government health-care system for low-income families.

¶ Included only in the model predicting multiple recurrences.

## References

1.
Bergeron MG. Treatment of pyelonephritis in adults.
Med Clin North Am

1995
;
79
:
619
–46.
2.
Foxman B, Klemstine KL, Brown PD. Acute pyelonephritis in US hospitals in 1997: hospitalization and in-hospital mortality.
Ann Epidemiol

2003
;
13
:
144
–50.
3.
Talan DA, Stamm WE, Hooton TM, et al. Comparison of ciprofloxacin (7 days) and trimethoprim-sulfamethoxazole (14 days) for acute uncomplicated pyelonephritis in women: a randomized trial.
JAMA

2000
;
283
:
1583
–90.
4.
Tenner SM, Yadven MW, Kimmel PL. Acute pyelonephritis: preventing complications through prompt diagnosis and proper therapy.

1992
;
91
:
261
–8.
5.
Parker J, Kunin C. Pyelonephritis in young women: a 10- to 20-year follow-up.
JAMA

1973
;
224
:
585
–90.
6.
Nicolle LE, Friesen D, Harding GKM, et al. Hospitalization for acute pyelonephritis in Manitoba, Canada, during the period from 1989 to 1992: impact of diabetes, pregnancy, and aboriginal origin.
Clin Infect Dis

1996
;
22
:
1051
–6.
7.
Raz R, Sakran W, Chazan B, et al. Long-term follow-up of women hospitalized for acute pyelonephritis.
Clin Infect Dis

2003
;
37
:
1014
–20.
8.
Institute of National Health Insurance. 1998. The statistics of health insurance. Seoul, South Korea: Jin Sung Press, 1999.
9.
Estimated population by age. Korea National Statistical Office (Daejeon, South Korea), 2003. (http://www.nso.go.kr/eng).
10.
Health insurance covered population by year. National Health Insurance Corporation, Seoul, South Korea, 2003. (http://www.nhic.or.kr/wbm/wbmb/index.html) (in Korean).
11.
Rice JA. Mathematical statistics and data analysis. 2nd ed. Belmont, CA: Duxbury Press, 1995:316–18.
12.
Therneau TM, Grambsch PM. Modeling survival data: extending the Cox model. New York, NY: Springer, 2000.
13.
The R project for statistical computing. Center for Computational Intelligence, Vienna University of Technology, Vienna, Austria, 2004. (http://www.r-project.org).
14.
Foxman B, Brown P. Epidemiology of urinary tract infections: transmission and risk factors, incidence, and costs.
Infect Dis Clin North Am

2003
;
17
:
227
–41.
15.
Wullt B, Bergsten G, Fischer H, et al. The host response to urinary tract infection.
Infect Dis Clin North Am

2003
;
17
:
279
–301.
16.
Parker AS, Cerhan JR, Lynch CF, et al. History of urinary tract infection and risk of renal cell carcinoma.
Am J Epidemiol

2004
;
159
:
42
–8.
17.
Dowell SF. Seasonal variation in host susceptibility and cycles of certain infectious disease.
Emerg Infect Dis

2001
;
7
:
369
–74.
18.
Pead L, Maskell R, Morris J. Staphylococcus saprophyticus as a urinary pathogen: a six year prospective survey.
BMJ (Clin Res Ed)

1985
;
291
:
1157
–9.
19.
Warren JW, Abrutyn E, Hebel JR, et al. Guidelines for antimicrobial treatment of uncomplicated acute bacterial cystitis and acute pyelonephritis in women.
Clin Infect Dis

1999
;
29
:
745
–58.
20.
Ki M, Park T, Yi SG, et al. Risk analysis of aseptic meningitis after measles-mumps-rubella vaccination in Korean children by using a case-crossover design.
Am J Epidemiol

2003
;
157
:
158
–65.
21.
2001 year book of health insurance corporation. Health Insurance Review Agency, Seoul, South Korea, 2004. (http://www.hira.or.kr) (in Korean).
22.
2002 health insurance review and evaluation statistical year book. Health Insurance Review Agency, Seoul, Korea, 2004. (http://www.hira.or.kr) (in Korean).