National Inpatient Sample (NIS) read program

nis2016_hospital_read.R

# this loads the 2016 NIS fixed width (asc) files into R
# it also saves the result as an fst file for much faster re-reading into R

library(data.table)
library(readr)
library(fst)

# load core data --------------------------------------------------------------------

nis_specs <- fread("./docs/nis_specs_core.csv")
nis_specs[, width := end - start + 1]
nis_specs[, varname := tolower(varname)]
nis_specs[, type := ifelse(type == "int", "i", ifelse(type %in% c("byte", "double", "long"), "d", "c"))]

missing_values <- as.character(quote(c(-99, -88, -66, -99.9999999, -88.8888888, -66.6666666, -9, -8, -6, -5, -9999, -8888, -6666, -999999999, -888888888, -666666666,-999, -888, -666)))

dt <- 
    read_fwf(
        "~/Dropbox (Outcomes Insights)/Data repository/HCUP/NIS_2016/NIS_2016_Core.ASC", 
        col_positions = fwf_widths(nis_specs$width), 
        col_types = paste0(nis_specs$type, collapse = ""),
        trim_ws = TRUE,
        na = missing_values
    )
setDT(dt)
setnames(dt, names(dt), nis_specs$varname)

write_fst(dt, "./data/analysis/hospital.fst", compress = 75)

dt <- read_fst("./data/analysis/hospital.fst", as.data.table = TRUE)

write_fst(dt, "./data/analysis/core.fst", compress = 100)

# load hospital data ----------------------------------------------------------------

nis_specs <- fread("./docs/nis_specs_hospital.csv")
nis_specs[, width := end - start + 1]
nis_specs[, varname := tolower(varname)]
nis_specs[, type := ifelse(type == "int", "i", ifelse(type %in% c("byte", "double", "long"), "d", "c"))]

missing_values <- as.character(quote(c(-99, -88, -66, -99.9999999, -88.8888888, -66.6666666, -9, -8, -6, -5, -9999, -8888, -6666, -999999999, -888888888, -666666666,-999, -888, -666)))

dt <- 
    read_fwf(
        "~/Dropbox (Outcomes Insights)/Data repository/HCUP/NIS_2016/NIS_2016_Hospital.ASC", 
        col_positions = fwf_widths(nis_specs$width), 
        col_types = paste0(nis_specs$type, collapse = ""),
        trim_ws = TRUE,
        # n_max = 1000,
        na = missing_values
    )
setDT(dt)
setnames(dt, names(dt), nis_specs$varname)

write_fst(dt, "./data/analysis/hospital.fst", compress = 100)

# load severity data ----------------------------------------------------------------

nis_specs <- fread("./docs/nis_specs_severity.csv")
nis_specs[, width := end - start + 1]
nis_specs[, varname := tolower(varname)]
nis_specs[, type := ifelse(type == "int", "i", ifelse(type %in% c("byte", "double", "long"), "d", "c"))]

missing_values <- as.character(quote(c(-99, -88, -66, -99.9999999, -88.8888888, -66.6666666, -9, -8, -6, -5, -9999, -8888, -6666, -999999999, -888888888, -666666666,-999, -888, -666)))

dt <- 
    read_fwf(
        "~/Dropbox (Outcomes Insights)/Data repository/HCUP/NIS_2016/NIS_2016_Severity.ASC", 
        col_positions = fwf_widths(nis_specs$width), 
        col_types = paste0(nis_specs$type, collapse = ""),
        trim_ws = TRUE,
        # n_max = 1000,
        na = missing_values
    )
setDT(dt)
setnames(dt, names(dt), nis_specs$varname)

write_fst(dt, "./data/analysis/severity.fst", compress = 100)


# copy the information below into a text editor and save as "nis_specs_xxx.csv" where "xxx" is
# "core", "hospital", or "severity" (see above)
# the information is taken from the Stata load programs
# white space is stripped when read into R, but kept below for legibility

# core data csv
type, varname, start, end
int,    AGE,                          1,   3   
byte,   AGE_NEONATE,                  4,   5   
byte,   AMONTH,                       6,   7   
byte,   AWEEKEND,                     8,   9   
byte,   DIED,                        10,  11   
double, DISCWT,                      12,  22   
byte,   DISPUNIFORM,                 23,  24   
byte,   DQTR,                        25,  26   
int,    DRG,                         27,  29   
byte,   DRGVER,                      30,  31   
int,    DRG_NoPOA,                   32,  34   
byte,   DXVER,                       35,  36   
byte,   ELECTIVE,                    37,  38   
byte,   FEMALE,                      39,  40   
int,    HCUP_ED,                     41,  43   
byte,   HOSP_DIVISION,               44,  45   
long,   HOSP_NIS,                    46,  50   
str,    I10_DX1,                     51,  57   
str,    I10_DX2,                     58,  64   
str,    I10_DX3,                     65,  71   
str,    I10_DX4,                     72,  78   
str,    I10_DX5,                     79,  85   
str,    I10_DX6,                     86,  92   
str,    I10_DX7,                     93,  99   
str,    I10_DX8,                    100, 106   
str,    I10_DX9,                    107, 113   
str,    I10_DX10,                   114, 120   
str,    I10_DX11,                   121, 127   
str,    I10_DX12,                   128, 134   
str,    I10_DX13,                   135, 141   
str,    I10_DX14,                   142, 148   
str,    I10_DX15,                   149, 155   
str,    I10_DX16,                   156, 162   
str,    I10_DX17,                   163, 169   
str,    I10_DX18,                   170, 176   
str,    I10_DX19,                   177, 183   
str,    I10_DX20,                   184, 190   
str,    I10_DX21,                   191, 197   
str,    I10_DX22,                   198, 204   
str,    I10_DX23,                   205, 211   
str,    I10_DX24,                   212, 218   
str,    I10_DX25,                   219, 225   
str,    I10_DX26,                   226, 232   
str,    I10_DX27,                   233, 239   
str,    I10_DX28,                   240, 246   
str,    I10_DX29,                   247, 253   
str,    I10_DX30,                   254, 260   
str,    I10_ECAUSE1,                261, 267   
str,    I10_ECAUSE2,                268, 274   
str,    I10_ECAUSE3,                275, 281   
str,    I10_ECAUSE4,                282, 288   
byte,   I10_NDX,                    289, 290   
int,    I10_NECAUSE,                291, 293   
byte,   I10_NPR,                    294, 295   
str,    I10_PR1,                    296, 302   
str,    I10_PR2,                    303, 309   
str,    I10_PR3,                    310, 316   
str,    I10_PR4,                    317, 323   
str,    I10_PR5,                    324, 330   
str,    I10_PR6,                    331, 337   
str,    I10_PR7,                    338, 344   
str,    I10_PR8,                    345, 351   
str,    I10_PR9,                    352, 358   
str,    I10_PR10,                   359, 365   
str,    I10_PR11,                   366, 372   
str,    I10_PR12,                   373, 379   
str,    I10_PR13,                   380, 386   
str,    I10_PR14,                   387, 393   
str,    I10_PR15,                   394, 400   
double, KEY_NIS,                    401, 410   
long,   LOS,                        411, 415   
byte,   MDC,                        416, 417   
byte,   MDC_NoPOA,                  418, 419   
int,    NIS_STRATUM,                420, 423   
byte,   PAY1,                       424, 425   
int,    PL_NCHS,                    426, 428   
int,    PRDAY1,                     429, 431   
int,    PRDAY2,                     432, 434   
int,    PRDAY3,                     435, 437   
int,    PRDAY4,                     438, 440   
int,    PRDAY5,                     441, 443   
int,    PRDAY6,                     444, 446   
int,    PRDAY7,                     447, 449   
int,    PRDAY8,                     450, 452   
int,    PRDAY9,                     453, 455   
int,    PRDAY10,                    456, 458   
int,    PRDAY11,                    459, 461   
int,    PRDAY12,                    462, 464   
int,    PRDAY13,                    465, 467   
int,    PRDAY14,                    468, 470   
int,    PRDAY15,                    471, 473   
byte,   PRVER,                      474, 475   
byte,   RACE,                       476, 477   
double, TOTCHG,                     478, 487   
byte,   TRAN_IN,                    488, 489   
byte,   TRAN_OUT,                   490, 491   
int,    YEAR,                       492, 495   
byte,   ZIPINC_QRTL,                496, 497

# hospital csv
type,varname,start,end
double, DISCWT,     1,11
byte, HOSP_BEDSIZE, 12,13
byte, HOSP_DIVISION,14,15
byte, HOSP_LOCTEACH,16,17
long, HOSP_NIS,     18,22
byte, HOSP_REGION,  23,24
byte, H_CONTRL,     25,26
int,  NIS_STRATUM,  27,30
long, N_DISC_U,     31,38
int,  N_HOSP_U,     39,42
long, S_DISC_U,     43,50
int,  S_HOSP_U,     51,54
long, TOTAL_DISC,   55,60
int,  YEAR,         61,64

# severity csv
type,varname,start,end
long,   HOSP_NIS,               1, 5
double, KEY_NIS,                6,15
int,    APRDRG,                16,19
byte,   APRDRG_Risk_Mortality, 20,21
byte,   APRDRG_Severity,       22,23

Do you have a load program for the state data (e.g., SID, SEDD)? Or have tips on how to create a load program for those datasets?

Sorry, I don't have any recent experience with those. I typically read things in by adapting the SAS or Stata load programs, if they are available. I do it manually or with a script depending on the complexity.

How did you get your specs data in a .csv format? Did you convert it from the NIS File Specifications website? https://hcup-us.ahrq.gov/db/nation/nis/nisfilespecs.jsp#2016NIS

As i mentioned above, I simply edit the available SAS or Stata load programs using a text editor like Sublime text that allows you to edit multiple rows simultaneously.

Thanks for sharing this! I adapted your code with a function to read specs from the stata load files online. I've tested 2016-2020 and it seems to work well https://gist.github.com/stephenrho/63f114281ee124cc6eec1db2433c9be5

@stephenrho -- that sounds great! Thanks for doing that.