clean1.Rmd

---
title: 'Aortic dissection: Basic survey analysis and results'
output:
  prettydoc::html_pretty:
  highlight: vignette
  theme: hpstr
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

This is the document for further cleaning the database to get variables into order. The data consists of years 2008 - 2015. Patients: All admissions for acute emergent aortic dissection. Cohort selection: From the earlier $rmarkdown$ file

```{r get the dataframe into the system}

library(haven);library(tidyverse);library(mosaic);library(survey)
library(skimr);library(srvyr);library(forcats)

df <- read_csv("G:/aor_dis/df/n0815.csv")


```


```{r create more var}

# aortic dissection surgical repair

rep <- c("3844","3845","3834","3835","3957","3958")

df$rep <-  with(df, ifelse((pr1 %in% rep | pr2 %in% rep | pr3 %in% rep),"yes","no"))

df %>% count(rep)



# also determine those patients who had associated cardiac surgical procedure 

cs <- c("3961","3963","3520","3521","3522","3600","3611","3612","3613","3614","3615","3616","3617","3918","3619", "3710","3711","3712","3832","3733","3734","3736")
df$cs <-  with(df, ifelse((pr1 %in% cs | pr2 %in% cs | pr3 %in% cs | pr4 %in% cs | pr5 %in% cs | pr6 %in% cs | pr7 %in% cs | pr8 %in% cs| pr9 %in% cs | pr10 %in% cs),"yes","no"))

df %>% count(cs)


# Determine those patients who have type A aortic dissection surgical repair

df$tasr <- with(df,ifelse((df$rep == "yes" & df$cs == "yes"),"yes","no"))

df %>% count(tasr)


# determine patients undergoing tevar

tevar <- c("3973")

df$tevar <-  with(df, ifelse((pr1 %in% tevar | pr2 %in% tevar | pr3 %in% tevar | pr4 %in% tevar | pr5 %in% tevar | pr6 %in% tevar | pr7 %in% tevar | pr8 %in% tevar| pr9 %in% tevar | pr10 %in% tevar),"yes","no"))

df %>% count(tevar)


# type B surgical repair patients

df$tbsr <- with(df, ifelse((df$rep == "yes" & df$cs == "no"), "yes","no"))

df %>% count(tbsr)

df$deal[df$tasr == "yes"] <- "tasr"
df$deal[df$tbsr == "yes"]<- "tbsr"
df$deal[df$tevar == "yes"]<- "tevar"
df$deal[df$tasr == "no" & df$tbsr == "no" & df$tevar == "no"]<- "mm"

df %>% count(deal)

df$comb <- with(df, ifelse((tasr == "yes" & tevar == "yes"),"yes","no"))

df %>% count(comb)

# race converted to new variable race_n

df <- df %>% mutate(race_n = fct_recode(factor(race),
                           "caucasian" = "1",
                           "AA" = "2",
                           "Hispanic"= "3",
                           "Others" = "4",
                           "Others" = "5",
                           "Others" = "6"))

df %>% count(race_n)

# type of hospital according to location and academic/non-academic

df <- df %>% mutate(hosp_type = fct_recode(factor(hosp_locteach),
                                           "urban/acad" = "3",
                                           "urban/nonacad" = "2",
                                           "rural" = "1"))

df %>% count(hosp_type)

# hospital size according to # of beds --- limited to 2015; not beyond

df <- df %>% mutate(hosp_size = fct_recode(factor(hosp_bedsize),
                                           
                                           "large" = "3",
                                           "medium" = "2",
                                           "small" = "1"))

df %>% count(hosp_size)

df <- df %>% mutate(insu = fct_recode(factor(pay1),
                                    "medicare" = "1",
                                    "medicaid" = "2",
                                    "private" = "3",
                                    "self-pay" = "4",
                                    "others" = "5",
                                    "others" = "6"
                                    ))
```

```{r skim to get basic understanding of dataframe after all changes and variable inserts}

glimpse(df)

```


Now, get survey design object created to start doing basic survey analysis:

```{r classical survey object}

df$count[df$year.x == 2008]<- 2619
df$count[df$year.x == 2009]<- 2751
df$count[df$year.x == 2010]<- 2504
df$count[df$year.x == 2011]<- 2777
df$count[df$year.x == 2012]<- 2644
df$count[df$year.x == 2013]<- 2754
df$count[df$year.x == 2014]<- 3044
df$count[df$year.x == 2015]<- 2355

df %>% count(count)
options(survey.lonely.psu = "certainty")

df_s<- svydesign(ids = ~hosp_nis, weights = ~discwt.x, strata = ~nis_stratum.x, data = df, nest = TRUE)

summary(df_s)
```

Now, using the survey design object *df_s* get standard estimates:

```{r}

# patients stratified for each therapy according to year

svytable(~deal+year.x , design = df_s)

# patients having combined type A aortic dissection repair and TEVAR during the same admission

svytable(~comb + year.x , design = df_s)
```


Create new survey object for tidy modelling

```{r srvyr survey object}

svy_df <- df %>% srvyr::as_survey_design(weights = discwt.x, strata = nis_stratum.x, id = hosp_nis, nest = TRUE)

svy_df %>%
  group_by(year.x) %>%
  summarise(age = srvyr::survey_mean(age, vartype = "ci"))

# contingency table for year and treatment

svytable(~year.x + deal, design = df_s)

df %>% count(year.x)

survey::svytable(~year.x + deal, design = svy_df)

```


Survey analysis for baseline demographics of the data:

```{r}
# create basic demographics and presentation of patients according to year and overall.

# overall demogrphics:

# age:

svymean(~age, design = df_s)

# aweekend:

svytable(~aweekend, design = df_s)

# female:

svytable(~female, design = df_s)

# region

svytable(~hosp_region, design = df_s)

# hosp_type

svytable(~hosp_type, design = df_s)

# hospital_size

svytable(~hosp_size, design = df_s)


                                  
```
To get the baseline $comorbidites-CM$ variables in the data, we will need to left_join the dataframes according to year.

We will need to first use $filter$ to get data according to year and then left_join data according to $nis-stratum$. This will need to be done for each year and then $rbind$ again to get the original dataframe back again.

Year 2008:

```{r merge and get year 2008 done}

# year 2008

n08 <- df %>% filter(year.x == 2008)

# get the original core file from that year

s08 <- read_stata("G:/NIS_DATA/nis2008/NIS_2008_Severity.dta")

names(s08) <- tolower(names(s08))

# then we will only select "cm" variables and "nis_stratum".

s08$key_nis <- s08$key

s08_s <- s08 %>% select(key_nis, cm_aids:cm_wghtloss)

m08 <- left_join(n08, s08_s, by = "key_nis")

glimpse(m08)
```

Year 2009:

```{r}

# year 2009

n09 <- df %>% filter(year.x == 2009)

# get the original core file from that year

s09 <- read_stata("G:/NIS_2009_Severity.dta")

names(s09) <- tolower(names(s09))

# then we will only select "cm" variables and "nis_stratum".

s09$key_nis <- s09$key

s09_s <- s09 %>% select(key_nis, cm_aids:cm_wghtloss)

m09 <- left_join(n09, s09_s, by = "key_nis")

glimpse(m09)
```

Year 2010:

```{r year 2010}

# year 2010

n10 <- df %>% filter(year.x == 2010)

# get the original core file from that year

s10 <- read_sas("G:/nis_2010_severity.sas7bdat")

names(s10) <- tolower(names(s10))

# then we will only select "cm" variables and "nis_stratum".

s10$key_nis <- s10$key

s10_s <- s10 %>% select(key_nis, cm_aids:cm_wghtloss)

m10 <- left_join(n10, s10_s, by = "key_nis")

glimpse(m10)

```

Year 2011:

```{r year 2011}

# year 2011

n11 <- df %>% filter(year.x == 2011)

# get the original core file from that year

s11 <- read_sas("G:/nis_2011_severity.sas7bdat")

names(s11) <- tolower(names(s11))

# then we will only select "cm" variables and "nis_stratum".

s11$key_nis <- s11$key

s11_s <- s11 %>% select(key_nis, cm_aids:cm_wghtloss)

m11 <- left_join(n11, s11_s, by = "key_nis")

glimpse(m11)

```

Year 2012:

```{r year 2012}

# year 2011

n12 <- df %>% filter(year.x == 2012)

# get the original core file from that year

s12 <- read_stata("G:/NIS_2012_Severity.dta")

names(s12) <- tolower(names(s12))

# then we will only select "cm" variables and "nis_stratum".


s12_s <- s12 %>% select(key_nis, cm_aids:cm_wghtloss)

m12 <- left_join(n12, s12_s, by = "key_nis")

glimpse(m12)

```


Year 2013:

```{r year 2013}

# year 2013

n13 <- df %>% filter(year.x == 2013)

# get the original core file from that year

s13 <- read_stata("G:/NIS_DATA/nis2013/NIS_2013_Severity.dta")

names(s13) <- tolower(names(s13))

# then we will only select "cm" variables and "nis_stratum".


s13_s <- s13 %>% select(key_nis, cm_aids:cm_wghtloss)

m13 <- left_join(n13, s13_s, by = "key_nis")

glimpse(m13)

```


Year 2014:

```{r year 2014}

# year 2014

n14 <- df %>% filter(year.x == 2014)

# get the original core file from that year

s14 <- read_stata("G:/NIS_2014_Severity.dta")

names(s14) <- tolower(names(s14))

# then we will only select "cm" variables and "nis_stratum".


s14_s <- s14 %>% select(key_nis, cm_aids:cm_wghtloss)

m14 <- left_join(n14, s14_s, by = "key_nis")

glimpse(m14)

```


Year 2015:

```{r year 2015}

# year 2015

n15 <- df %>% filter(year.x == 2015)

# get the original core file from that year

s15 <- read_stata("G:/NIS_2015Q1Q3_Severity.dta")

names(s15) <- tolower(names(s15))

# then we will only select "cm" variables and "nis_stratum".


s15_s <- s15 %>% select(key_nis, cm_aids:cm_wghtloss)

m15 <- left_join(n15, s15_s, by = "key_nis")

glimpse(m15)

```

*rbind* all dataframes together again:

```{r rbind all dataframes}

df_new <- rbind(m08,m09,m10,m11,m12,m13,m14,m15)

```

Create new standard and tidy survey object again so that we can do more survey analysis.

```{r new survey object}

options(survey.lonely.psu = "certainty")

dfnew_s<- svydesign(ids = ~hosp_nis, weights = ~discwt.x, strata = ~nis_stratum.x, data = df_new, nest = TRUE)

summary(dfnew_s)

# modified survey object for tidy analysis

svy_dfnew <- df_new %>% srvyr::as_survey_design(weights = discwt.x, strata = nis_stratum.x, id = hosp_nis, nest = TRUE)
```

Now the dataframe is almost complete. Plan to decide whether to focus the paper on all patients with acute thoracic dissection or only those who underwent type a aortic dissection repair. 

For now I will plan to subset the data for only type a dissection repair patients and do the analysis for these patients.
1. Trend analysis of prevalence for repair of type a
2. demographics overall and year wise trends esp. for important variables like age, COPD, gender, renal dysfunction
3. Identify variables important for type a aortic dissection from icd9 codes: marfans syndrome, medial cystic necrosis and any connective tissue disorders that i can identify using icd9 codes. then proceed to provide overall and trends for each predisposing factor.
4. clinical outcome: los, mortality, stroke, acute renal failure, respiratory failure, discharge planning and discharge location

marfan syndrome = 759.82

thoracic aortic ectasia = 447.71

turners syndrome = 758.6

ehler danlos syndrome = 756.83




marfan syndrome:

```{r further important diagnoses1}

marf <- as.character(c(75982))

df_new$marfan <- with(df_new, ifelse((dx1 %in% marf | dx2 %in% marf | dx3 %in% marf | dx4 %in% marf | dx5  %in% marf | 
                                        dx6 %in% marf | dx7 %in% marf | dx8 %in% marf | dx9 %in% marf | dx10 %in% marf), 1, 0))

df_new %>% count(marfan)

```

ehler danlos syndrome:

```{r further important diagnoses2}

ehd <- as.character(c(75683))

df_new$ehd <- with(df_new, ifelse((dx1 %in% ehd | dx2 %in% ehd | dx3 %in% ehd | dx4 %in% ehd | dx5  %in% ehd | 
                                        dx6 %in% ehd | dx7 %in% ehd | dx8 %in% ehd | dx9 %in% ehd | dx10 %in% ehd), 1, 0))

df_new %>% count(ehd)

```

thoracic aortic ectasia:

```{r}

aae <- as.character(c(44771))

df_new$aae <- with(df_new, ifelse((dx1 %in% aae | dx2 %in% aae | dx3 %in% aae | dx4 %in% aae | dx5  %in% aae | 
                                        dx6 %in% aae | dx7 %in% aae | dx8 %in% aae | dx9 %in% aae | dx10 %in% aae), 1, 0))

df_new %>% count(aae)
```


Now I want to focus first only on patients with type a aortic dissection

```{r dataframe for only type a aortic dissection}

ta <- df_new %>% filter(tasr == "yes")

dim(ta)
```

Survey methods for type a aortic dissection repair patients:

```{r}



# new design object

ta_s <- ta %>% srvyr::as_survey_design(weights = discwt.x, strata = nis_stratum.x, id = hosp_nis, nest = TRUE)


ta_s %>%
  summarise(age = survey_mean(age,NA.RM = TRUE, vartype = c("ci"), df = Inf))


ta_s %>%
  group_by(year.x) %>%
  summarise(age = survey_mean(age, na.rm = TRUE, vartype = c("ci"), df = Inf))


# gender overall

ta_s %>%
  summarise(female = survey_total(female, na.rm = TRUE, vartype = c("ci"), df = Inf))

# gender_year

ta_s %>%
  group_by(year.x) %>%
  summarise(female = survey_total(female, na.rm = TRUE, vartype = c("ci"), df = Inf))

# type of hospital

# overall for all type a aortic dissection repairs

survey::svytable(~hosp_type, design = ta_s)

# hospital type by year 

survey::svytable(~hosp_type + year.x, design = ta_s)
```