major strength of . From this perspective, the low values of found with high sensitivity and specificity and low preva¬ lence are analagous to the low predictive values of diagnostic
tests found under the same conditions. Feinstein and Cicchetti4·5 further explored the relationship between and prevalence in a highly detailed and seemingly definitive investigation of this and a second apparent paradox of . They first showed that the most complete character¬ ization of the degree of consistency between two judges is through the use of two measures, positive and negative agree¬ ment, which are analagous to sensitivity and specificity. They then showed that "the paradoxes of are desirable since they appropriately 'penalize' inequalities in positive and negative agreement." Feinstein and Cicchetti concluded that is the preferred single index for quantifying the magnitude of agree¬ ment and comparing studies of observer variability: "Its ap¬ parent paradoxes are the inevitable and desirable result of a
correction factor that helps adjust or 'standardize' disparities in the populations under study and in performance of the
observers."4,5
Finally, Berry's estimate of the prevalence of inadequate in the studies I reviewed is too low. He provides values of based on prevalences of 0.01 to 0.10. Of the 12 studies included in my review, three did not report the frequency with which the judges used the different possible ratings. There were 19 coefficients computed in the other nine studies; for none was there a frequency of less than 0.10 for the rating indicating inadequate care. Thus, even if one agreed with Berry that was a flawed measure of agreement when prevalence is highly imbalanced, this deficiency would have care
little relevance to the studies included in my review. In con¬ clusion, Berry's letter raises important issues, but is
misleading.
Ronald L. Goldman, PhD, MPH Washington, DC 1. Kraemer HC. Ramifications of a population model for \g=k\as a coefficient of reliability. Psychometrika. 1979;44:461-472. 2. Grove WM, Andreasen NC, McDonald-Scott P, Keller MB, Shapiro WB. Reliability studies of psychiatric practice: theory and practice. Arch Gen Psychiatry. 1981;
38:408-413. 3. Shrout PE, Spitzer RL, Fleiss JL. Quantification of agreement in psychiatric diagnosis revisited. Arch Gen Psychiatry. 1987;44:172-177. 4. Feinstein AR, Cicchetti DV. High agreement but low kappa, I: the problem of two paradoxes. J Clin Epidemiol. 1990;43:543-550. 5. Cicchetti DV, Feinstein AR. High agreement but low kappa, II: resolving the paradoxes. J Clin Epidemiol. 1990;43:551-558.
The Los Angeles Riots of 2014: How to Avert Them To the Editor.\p=m-\Inthe aftermath of the Los Angeles riots of 1992, there is a discomforting sense that American society has embarked on an inevitable, repetitive cycle. Although shocked, can communities across the nation truly be surprised by the events that transpired or by their magnitude? A ritualized laying of blame, promise of federal aid to cities, and the commission investigations are all unfolding predictably. Of concern is that fatigued, well-worn rationales continue to guide the policy debate and response. These can be summarized as advocacy of education to engender economic opportunity and escape from poverty. Absent from this discussion is any focus on physical or mental health. As such, this formula will not prevent repetition of riots in Los Angeles and other American cities. While the social, political, and economic roots of poverty and community dysfunction must certainly be addressed, they do not exist in a vacuum. As long as the public health status of the inner-city poor remains undiscussed and unattended at the center of efforts to rebuild and reshape these communities, there can be no successful prevention of such tragic episodes. Substance abuse is es¬ sentially a public health problem. Unwanted pregnancy and intentional injury caused by domestic and youth violence are
public health problems. In the simplest terms, a malnourished brain cannot be educated for economic opportunity. A nutri¬ tionally weakened child with recurrent absences from school for avoidable respiratory tract and other infections is unlikely
poverty. Contrary to public perception, the riots were not perpe¬ trated predominantly by blacks. Of 7066 individuals arrested for looting in the city of Los Angeles in the first week of rioting, less than 40% were black and many of the rest were Hispanic.1 But as long as black life expectancy falls short of that of whites by a half decade,2 and as long as black ado¬ lescent pregnancy and black infant mortality rates are twice as high as that of whites,3 the Los Angeles riots may recur. Similar data can be cited to highlight serious Hispanic health problems and inequities.4 There is no great mystery under¬ lying this problem. Organized medicine cannot ignore the fact that sustained socioeconomic polarization has given at least equal import to the biological basis of enfranchising individ¬ uals as to the provision of social and economic opportunity. Our profession has an essential role in bringing to bear all that public health and medicine have in their armamentarium— and there is much with known, proven efficacy across a broad range of so-called social problems— to ensure that the human to escape
substrate and necessary prerequisites for social advance¬ ment are intact. Only then can the Los Angeles riots of 2014 be averted with certainty. George A. Gellert, MD, MPH, MPA
Orange County Health Care Agency Santa Ana, Calif
1.
Monthly Arrest Statistics. Los Angeles, Calif: City of Los Angeles Police Depart-
ment; April 1992. 2.
Closing the Gap: Health and Minorities in the United States. Washington, DC: Dept of Health and Human Services, Public Health Service, Office of Minority
US
Health;
1988.
Report of the Secretary's Task Force on Black and Minority Health. Washington, DC: US Dept of Health and Human Services; 1985. 4. Council on Scientific Affairs, American Medical Association. Hispanic health in the 3.
United States. JAMA. 1991;265:248-252.
Transparent Polyurethane Film as a Catheter Dressing To the Editor.\p=m-\Ina recent article, Hoffmann et al1 found statistically significant increases in catheter-tip infection rates for transparent as opposed to gauze dressings. With hundreds of patients in the combined studies, they found no significant increase in phlebitis, infiltration or skin colonization for peripheral venous catheters, and none in bacteremia or sepsis for central venous catheters. Unfortunately, they conclude, "While it should not be assumed that transparent dressings would increase bacteremia, it is likely that higher rates of catheter-tip infection will result in an elevated rate of bacteremia." They say that the "literature [shows] that catheter-tip infection is an excellent harbinger of bacteremia and sepsis." They cite Maki,2 but his article says "positive cultures found using this technique have a 15 to 40 percent association with concomitant bacteremias." Also, Nahass and Weinstein3 estimate the positive predictive value to be 17%. "Excellent harbinger" is hardly appropriate. Actually, this relationship is unimportant because the authors have wisely analyzed clinical infection and find no statistically significant increase in either bacte¬ remia or sepsis. The authors take license in calling something a trend: "[T]he results did not reach statistical significance, but the trend suggests that transparent dressings adversely effect the risk of infection." This mocks statistical conclusions. Are they suggesting something about the responses of the next several hundred patients? This statement is not supported by data and merely reflects the authors' biases. Many of the studies considered had important differences
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 06/10/2015
.
..
.
..
in the
transparent dressing group: longer dressing change interval, longer duration of catheter use, and less use of topical antibiotics. The authors discuss but dismiss these covariates. Also, they combine studies with different brands of transparent dressing—brands are not interchangeable.4 A typographical error in Table 3 inadvertently exchanged proportions of phlebitis in the Nicola and DeChairo study: 78/255 (30.6%) applies to gauze and not transparent; 58/270 (21.5%) applies to transparent and not gauze dressings (the relative risk [RR] is correct.) This difference is statistically significant in favor of transparent dressing. Maki and Will5 did not find skin colonization rates significantly greater using transparent as opposed to gauze dressings when both were changed every 2 days. Arterial catheter data from Maki and Will6 should have been excluded; some transparent dressings are contraindicated for arterial use. Abstracts are subject to the same biases as published articles, and so do not "avoid publication bias." Significant differences in catheter-tip infection do not im¬ ply correspondingly significant differences in clinical infec¬ tion. There were no statistically significant increases in any clinical infection due to transparent dressings. Finally, co¬ variates not considered in the authors' statistical analyses could more than make up for observed differences in dressing
type.
Donald A. Berry, PhD Duke University Durham, NC Since 1980, Dr Berry has been a paid consultant to the 3M manufacturer of Tegaderm® transparent dressing.
Corporation,
1. Hoffmann KK, Weber DJ, Samsa GP, Rutala WA. Transparent polyurethane film intravenous catheter dressing: a meta-analysis of the infection risks. JAMA.
as an
1992;267:2072-2076. therapy. In: Bennett JV, Brachman PS, eds. Hospital Infections. 2nd ed. Boston, Mass: Little Brown & Co Inc; 1986:561-580. 3. Nahass RG, Weinstein MP. Qualitative intravascular catheter tip cultures do not predict catheter-related bacteremia. Diagn Microbiol Infect Dis. 1990;13:223-226. 4. Young GP, Alexeyeff M, Russell DM, Thomas RJS. Catheter sepsis during parenteral nutrition: the safety of long-term OpSite dressings. JPEN J Parenter Enteral Nutr. 1988;12:365-370. 5. Maki DG, Will L. Colonization and infection associated with transparent dressings for central venous catheters: a comparative trial. In: Program and abstracts of the Association for Practitioners in Infection Control Educational Conference; June 3-7, 1984; Washington, DC. Abstract poster session, board 21. 6. Maki DG, Will L. Study of polyantibiotic and povidone-iodine ointments on central venous and arterial catheter-sites dressed with gauze or polyurethane dressing. In: Program and abstracts of the 26th Interscience Conference on Antimicrobial Agents and Chemotherapy; September 28-October 1, 1986; New Orleans, La. 2. Maki DG. Infections due to infusion
Abstract 1041.
Reply.\p=m-\Whencomparing the infection risks of transparvs gauze dressings used on central venous catheters, the RRs were 1.78 for catheter-tip infection, 1.63 for bacteremia, and 1.69 for catheter sepsis. These RRs represent the best In
ent
assessment of the overall risk associated with the
use
of
transparent dressings. (For catheter-tip infection, P
tests found under the same conditions. Feinstein and Cicchetti4·5 further explored the relationship between and prevalence in a highly detailed and seemingly definitive investigation of this and a second apparent paradox of . They first showed that the most complete character¬ ization of the degree of consistency between two judges is through the use of two measures, positive and negative agree¬ ment, which are analagous to sensitivity and specificity. They then showed that "the paradoxes of are desirable since they appropriately 'penalize' inequalities in positive and negative agreement." Feinstein and Cicchetti concluded that is the preferred single index for quantifying the magnitude of agree¬ ment and comparing studies of observer variability: "Its ap¬ parent paradoxes are the inevitable and desirable result of a
correction factor that helps adjust or 'standardize' disparities in the populations under study and in performance of the
observers."4,5
Finally, Berry's estimate of the prevalence of inadequate in the studies I reviewed is too low. He provides values of based on prevalences of 0.01 to 0.10. Of the 12 studies included in my review, three did not report the frequency with which the judges used the different possible ratings. There were 19 coefficients computed in the other nine studies; for none was there a frequency of less than 0.10 for the rating indicating inadequate care. Thus, even if one agreed with Berry that was a flawed measure of agreement when prevalence is highly imbalanced, this deficiency would have care
little relevance to the studies included in my review. In con¬ clusion, Berry's letter raises important issues, but is
misleading.
Ronald L. Goldman, PhD, MPH Washington, DC 1. Kraemer HC. Ramifications of a population model for \g=k\as a coefficient of reliability. Psychometrika. 1979;44:461-472. 2. Grove WM, Andreasen NC, McDonald-Scott P, Keller MB, Shapiro WB. Reliability studies of psychiatric practice: theory and practice. Arch Gen Psychiatry. 1981;
38:408-413. 3. Shrout PE, Spitzer RL, Fleiss JL. Quantification of agreement in psychiatric diagnosis revisited. Arch Gen Psychiatry. 1987;44:172-177. 4. Feinstein AR, Cicchetti DV. High agreement but low kappa, I: the problem of two paradoxes. J Clin Epidemiol. 1990;43:543-550. 5. Cicchetti DV, Feinstein AR. High agreement but low kappa, II: resolving the paradoxes. J Clin Epidemiol. 1990;43:551-558.
The Los Angeles Riots of 2014: How to Avert Them To the Editor.\p=m-\Inthe aftermath of the Los Angeles riots of 1992, there is a discomforting sense that American society has embarked on an inevitable, repetitive cycle. Although shocked, can communities across the nation truly be surprised by the events that transpired or by their magnitude? A ritualized laying of blame, promise of federal aid to cities, and the commission investigations are all unfolding predictably. Of concern is that fatigued, well-worn rationales continue to guide the policy debate and response. These can be summarized as advocacy of education to engender economic opportunity and escape from poverty. Absent from this discussion is any focus on physical or mental health. As such, this formula will not prevent repetition of riots in Los Angeles and other American cities. While the social, political, and economic roots of poverty and community dysfunction must certainly be addressed, they do not exist in a vacuum. As long as the public health status of the inner-city poor remains undiscussed and unattended at the center of efforts to rebuild and reshape these communities, there can be no successful prevention of such tragic episodes. Substance abuse is es¬ sentially a public health problem. Unwanted pregnancy and intentional injury caused by domestic and youth violence are
public health problems. In the simplest terms, a malnourished brain cannot be educated for economic opportunity. A nutri¬ tionally weakened child with recurrent absences from school for avoidable respiratory tract and other infections is unlikely
poverty. Contrary to public perception, the riots were not perpe¬ trated predominantly by blacks. Of 7066 individuals arrested for looting in the city of Los Angeles in the first week of rioting, less than 40% were black and many of the rest were Hispanic.1 But as long as black life expectancy falls short of that of whites by a half decade,2 and as long as black ado¬ lescent pregnancy and black infant mortality rates are twice as high as that of whites,3 the Los Angeles riots may recur. Similar data can be cited to highlight serious Hispanic health problems and inequities.4 There is no great mystery under¬ lying this problem. Organized medicine cannot ignore the fact that sustained socioeconomic polarization has given at least equal import to the biological basis of enfranchising individ¬ uals as to the provision of social and economic opportunity. Our profession has an essential role in bringing to bear all that public health and medicine have in their armamentarium— and there is much with known, proven efficacy across a broad range of so-called social problems— to ensure that the human to escape
substrate and necessary prerequisites for social advance¬ ment are intact. Only then can the Los Angeles riots of 2014 be averted with certainty. George A. Gellert, MD, MPH, MPA
Orange County Health Care Agency Santa Ana, Calif
1.
Monthly Arrest Statistics. Los Angeles, Calif: City of Los Angeles Police Depart-
ment; April 1992. 2.
Closing the Gap: Health and Minorities in the United States. Washington, DC: Dept of Health and Human Services, Public Health Service, Office of Minority
US
Health;
1988.
Report of the Secretary's Task Force on Black and Minority Health. Washington, DC: US Dept of Health and Human Services; 1985. 4. Council on Scientific Affairs, American Medical Association. Hispanic health in the 3.
United States. JAMA. 1991;265:248-252.
Transparent Polyurethane Film as a Catheter Dressing To the Editor.\p=m-\Ina recent article, Hoffmann et al1 found statistically significant increases in catheter-tip infection rates for transparent as opposed to gauze dressings. With hundreds of patients in the combined studies, they found no significant increase in phlebitis, infiltration or skin colonization for peripheral venous catheters, and none in bacteremia or sepsis for central venous catheters. Unfortunately, they conclude, "While it should not be assumed that transparent dressings would increase bacteremia, it is likely that higher rates of catheter-tip infection will result in an elevated rate of bacteremia." They say that the "literature [shows] that catheter-tip infection is an excellent harbinger of bacteremia and sepsis." They cite Maki,2 but his article says "positive cultures found using this technique have a 15 to 40 percent association with concomitant bacteremias." Also, Nahass and Weinstein3 estimate the positive predictive value to be 17%. "Excellent harbinger" is hardly appropriate. Actually, this relationship is unimportant because the authors have wisely analyzed clinical infection and find no statistically significant increase in either bacte¬ remia or sepsis. The authors take license in calling something a trend: "[T]he results did not reach statistical significance, but the trend suggests that transparent dressings adversely effect the risk of infection." This mocks statistical conclusions. Are they suggesting something about the responses of the next several hundred patients? This statement is not supported by data and merely reflects the authors' biases. Many of the studies considered had important differences
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 06/10/2015
.
..
.
..
in the
transparent dressing group: longer dressing change interval, longer duration of catheter use, and less use of topical antibiotics. The authors discuss but dismiss these covariates. Also, they combine studies with different brands of transparent dressing—brands are not interchangeable.4 A typographical error in Table 3 inadvertently exchanged proportions of phlebitis in the Nicola and DeChairo study: 78/255 (30.6%) applies to gauze and not transparent; 58/270 (21.5%) applies to transparent and not gauze dressings (the relative risk [RR] is correct.) This difference is statistically significant in favor of transparent dressing. Maki and Will5 did not find skin colonization rates significantly greater using transparent as opposed to gauze dressings when both were changed every 2 days. Arterial catheter data from Maki and Will6 should have been excluded; some transparent dressings are contraindicated for arterial use. Abstracts are subject to the same biases as published articles, and so do not "avoid publication bias." Significant differences in catheter-tip infection do not im¬ ply correspondingly significant differences in clinical infec¬ tion. There were no statistically significant increases in any clinical infection due to transparent dressings. Finally, co¬ variates not considered in the authors' statistical analyses could more than make up for observed differences in dressing
type.
Donald A. Berry, PhD Duke University Durham, NC Since 1980, Dr Berry has been a paid consultant to the 3M manufacturer of Tegaderm® transparent dressing.
Corporation,
1. Hoffmann KK, Weber DJ, Samsa GP, Rutala WA. Transparent polyurethane film intravenous catheter dressing: a meta-analysis of the infection risks. JAMA.
as an
1992;267:2072-2076. therapy. In: Bennett JV, Brachman PS, eds. Hospital Infections. 2nd ed. Boston, Mass: Little Brown & Co Inc; 1986:561-580. 3. Nahass RG, Weinstein MP. Qualitative intravascular catheter tip cultures do not predict catheter-related bacteremia. Diagn Microbiol Infect Dis. 1990;13:223-226. 4. Young GP, Alexeyeff M, Russell DM, Thomas RJS. Catheter sepsis during parenteral nutrition: the safety of long-term OpSite dressings. JPEN J Parenter Enteral Nutr. 1988;12:365-370. 5. Maki DG, Will L. Colonization and infection associated with transparent dressings for central venous catheters: a comparative trial. In: Program and abstracts of the Association for Practitioners in Infection Control Educational Conference; June 3-7, 1984; Washington, DC. Abstract poster session, board 21. 6. Maki DG, Will L. Study of polyantibiotic and povidone-iodine ointments on central venous and arterial catheter-sites dressed with gauze or polyurethane dressing. In: Program and abstracts of the 26th Interscience Conference on Antimicrobial Agents and Chemotherapy; September 28-October 1, 1986; New Orleans, La. 2. Maki DG. Infections due to infusion
Abstract 1041.
Reply.\p=m-\Whencomparing the infection risks of transparvs gauze dressings used on central venous catheters, the RRs were 1.78 for catheter-tip infection, 1.63 for bacteremia, and 1.69 for catheter sepsis. These RRs represent the best In
ent
assessment of the overall risk associated with the
use
of
transparent dressings. (For catheter-tip infection, P
